From e0e7bc573cea001ecf81ba5464de0b010552f910 Mon Sep 17 00:00:00 2001 From: lofyer Date: Tue, 11 Aug 2026 16:49:51 +0800 Subject: [PATCH] feat: add document OCR and offline model archives --- DESIGN.md | 23 + docs/features/doc-extract.md | 331 ++++++ electron.vite.config.ts | 3 + package-lock.json | 185 ++- package.json | 14 + src/main/context-manager.ts | 31 +- src/main/document-ocr-broker.test.ts | 53 + src/main/document-ocr-broker.ts | 131 +++ src/main/document-ocr-model-catalog.ts | 204 ++++ src/main/document-ocr-model-manager.test.ts | 341 ++++++ src/main/document-ocr-model-manager.ts | 899 +++++++++++++++ src/main/document-parsing-service.test.ts | 130 +++ src/main/document-parsing-service.ts | 270 +++++ .../document-parsing-settings-store.test.ts | 120 ++ src/main/document-parsing-settings-store.ts | 181 +++ src/main/index.ts | 36 +- src/main/ipc.test.ts | 181 ++- src/main/ipc.ts | 286 ++++- .../document-parser-electron-runtime.test.ts | 50 + src/main/knowledge/document-parser.ts | 55 +- src/main/knowledge/knowledge-service.ts | 25 +- src/main/model-archive.test.ts | 211 ++++ src/main/model-archive.ts | 617 ++++++++++ src/main/speech/speech-model-manager.test.ts | 41 + src/main/speech/speech-model-manager.ts | 143 +++ src/preload/index.ts | 106 +- src/preload/preload-sandbox.test.ts | 13 + .../DocumentParsingSettingsSection.test.tsx | 377 ++++++ .../src/DocumentParsingSettingsSection.tsx | 1017 +++++++++++++++++ src/renderer/src/SettingsPanel.test.tsx | 3 +- src/renderer/src/SettingsPanel.tsx | 6 + .../src/SpeechModelSettingsSection.test.tsx | 98 +- .../src/SpeechModelSettingsSection.tsx | 74 +- src/renderer/src/WorkspacePrimitives.tsx | 24 +- src/renderer/src/document-ocr-bridge.ts | 203 ++++ src/renderer/src/document-ocr-worker.ts | 233 ++++ .../src/i18n/locales/en-US/settings.ts | 177 ++- .../i18n/locales/en-US/settingsSections.ts | 13 +- .../src/i18n/locales/zh-CN/settings.ts | 163 ++- .../i18n/locales/zh-CN/settingsSections.ts | 13 +- src/renderer/src/main.tsx | 3 + src/renderer/src/settings-categories.ts | 4 + src/renderer/src/styles.css | 451 ++++++++ src/shared/contracts.ts | 46 +- src/shared/document-parsing-contracts.ts | 310 +++++ src/shared/ipc-channels.ts | 21 +- 46 files changed, 7831 insertions(+), 85 deletions(-) create mode 100644 docs/features/doc-extract.md create mode 100644 src/main/document-ocr-broker.test.ts create mode 100644 src/main/document-ocr-broker.ts create mode 100644 src/main/document-ocr-model-catalog.ts create mode 100644 src/main/document-ocr-model-manager.test.ts create mode 100644 src/main/document-ocr-model-manager.ts create mode 100644 src/main/document-parsing-service.test.ts create mode 100644 src/main/document-parsing-service.ts create mode 100644 src/main/document-parsing-settings-store.test.ts create mode 100644 src/main/document-parsing-settings-store.ts create mode 100644 src/main/knowledge/document-parser-electron-runtime.test.ts create mode 100644 src/main/model-archive.test.ts create mode 100644 src/main/model-archive.ts create mode 100644 src/renderer/src/DocumentParsingSettingsSection.test.tsx create mode 100644 src/renderer/src/DocumentParsingSettingsSection.tsx create mode 100644 src/renderer/src/document-ocr-bridge.ts create mode 100644 src/renderer/src/document-ocr-worker.ts create mode 100644 src/shared/document-parsing-contracts.ts diff --git a/DESIGN.md b/DESIGN.md index a8af97c..3cc3a6e 100644 --- a/DESIGN.md +++ b/DESIGN.md @@ -513,6 +513,26 @@ GoodBuddy 是可调整窗口大小的桌面应用。响应式设计优先保证 - 自动生效、仅执行即时命令或自行管理编辑流程的分类不显示全局保存操作。窄窗口下操作区可以换行，但保存入口必须保持清晰可见。 - 保存或测试成功统一进入应用通知视口，并按全局规则自动消失，不在分类页头或内容卡片中保留持久成功文案。加载、保存和测试错误显示在分类页头下方，并保留可处理的上下文。 +### 13.8 文档解析设置 + +- 设置中心新增独立的“文档解析”分类，统一管理聊天附件、知识库导入以及后续文档审阅场景使用的提取、转换和 OCR 策略。OCR 不作为普通对话模型出现在“模型连接”中。 +- 分类页头说明文档解析的跨场景作用，右侧依次显示“测试解析”和“保存设置”；保存位于最右侧。测试必须选择真实文件并执行实际解析，不能只检查模型文件或接口连通性。 +- 页面首先显示原生解析、文档转换和 OCR 的运行状态，并明确当前可处理格式、回退能力与不可用原因。部分能力未配置时使用“部分可用”状态，不得把原生文本解析一并标记为失败。 +- “使用场景”分别配置聊天附件和知识库导入。普通用户选择“自动解析”“快速文本”“完整索引”等预设；阈值、并发和超时放入默认折叠的高级设置。 +- 本地 OCR 的全平台基线使用同一组 PP-OCRv6 ONNX 模型和 ONNX Runtime WebAssembly，在 Windows、macOS、Linux 的 x64 与 arm64 上保持相同功能。原生 ONNX、WebGPU、DirectML、CoreML 或 CUDA 只能作为可选加速，失败时必须回退到 WASM CPU。 +- OCR 模型管理与语音模型保持一致：应用不内置权重，用户可按需从 ModelScope 下载，也可在联网设备导出 ZIP 并在离线或内网设备直接导入。语音和 OCR 模型的下载、取消、ZIP 导入、ZIP 导出、删除与打开受管目录使用同一交互语义；ZIP 操作不得隐式切换当前模型或保存解析设置。 +- OCR 模型卡片必须持续显示来源、语言、运行时、体积、安装状态和许可。“打开 ModelScope”直接位于卡片右上角，不再使用“模型详情与手动导入”折叠区。窄窗口下仓库操作换行到模型摘要下方，仍须保持可访问名称和键盘操作。 +- PP-OCRv6 提供三个已实现档位：Tiny 约 6 MiB，适合低资源设备；Small 约 30 MiB，官方支持 50 种语言并作为推荐档位；Medium 约 132 MiB，官方支持 50 种语言、质量更高但速度较慢，界面必须提示其更高的内存占用和延迟。 +- 本地模型按受管目录和固定清单加载。ModelScope 下载地址必须固定不可变 revision、字节数和 SHA-256；下载先进入临时目录，全部校验成功后再原子安装。识别时不得从网络或可变分支临时加载模型。 +- 模型 ZIP 使用版本化的 `goodbuddy-model.json` 清单，声明模型类型、内置目录 ID、文件角色、大小与 SHA-256。导出前重新校验已安装文件；导入时限制压缩包大小、条目数、单文件和总展开大小，拒绝路径穿越、重复、未知、缺失或嵌套条目，并以应用内置目录重新校验后原子安装。ZIP 内的自声明信息不能扩大受信任模型集合。 +- PDF 先读取文本层。仅当页面无有效文本、乱码比例过高或用户选择“始终 OCR”时渲染该页并识别；不得因为单页需要 OCR 而丢弃其他页面已经提取的可靠文本。 +- DOCX、XLSX、PPTX 优先保留段落、单元格、公式、备注等原生语义。转换为 PDF 用于补充版面、页码、图表和图片理解，不作为唯一中间格式。 +- DOC、XLS、PPT 等旧格式通过受控转换 Provider 生成新式 Office 文档和 PDF。转换子进程必须禁用宏和网络，限制输入、输出、内存、超时与临时目录，并在关闭或取消时清理。 +- OCR 来源使用“本地模型 / 远程服务”互斥选择。选择本地后显示模型下载、模型下拉选择和本地运行参数；选择远程后显示 MinerU、PaddleOCR-VL 等服务连接配置。未实现的远程服务入口保持可读但禁用，不再增加与来源选择重复的“隐私与云端处理”授权区。 +- 用户配置并保存远程 OCR 服务即表示选择该处理路径，不再逐场景重复询问。界面仍须明确显示当前服务名称、处理范围和远程属性，API 密钥只保存在主进程加密设置中，未选中远程服务时不得上传文档。 +- 解析结果使用统一文档结构，至少保留文档标题、来源格式、页码或工作表定位、正文块、置信度、处理方式和警告。聊天附件对结果做有界截断，知识库使用完整结果分块和索引。 +- 测试结果显示文件类型、页数、实际工作流、提取字数、OCR 页数、耗时和警告。测试文件不得自动进入聊天上下文或知识库。 + ## 14. 文案规则 - 使用简体中文，动词直接、对象明确。 @@ -562,6 +582,7 @@ GoodBuddy 是可调整窗口大小的桌面应用。响应式设计优先保证 - [ ] 智能心跳迁移到 `dashboard`，统一状态卡片、配置和运行历史层级。 - [ ] 任务迁移到 `standard`，活动记录迁移到 `dashboard`，统一导航、筛选和表格行为。 - [ ] 设置中心使用共享分类定义与 `SettingsCategoryHeader`，将保存与测试操作统一放到分类页头右侧，并把成功反馈接入应用通知。 +- [ ] 文档解析设置统一聊天附件与知识库的解析预设、OCR 状态、转换状态、隐私限制和真实文件测试。 ### 15.5 验收 @@ -572,6 +593,8 @@ GoodBuddy 是可调整窗口大小的桌面应用。响应式设计优先保证 - [ ] 验证页面范围、对象范围和操作范围在关键流程中始终可见。 - [ ] 验证删除、批量操作、停止运行和清空历史符合风险等级策略。 - [ ] 验证加载中、首次为空、筛选无结果、搜索无结果、失败和只读状态不会互相混用。 +- [ ] 在 Windows、macOS、Linux 的 x64 与 arm64 上执行真实本地 OCR，并验证 WASM CPU 回退、取消、超时和离线运行。 +- [ ] 在联网设备导出语音与 OCR 模型 ZIP，在离线设备导入后执行真实推理；验证错误模型 ID、篡改文件、路径穿越、未知条目和压缩炸弹均被拒绝。 ## 16. 完成标准 diff --git a/docs/features/doc-extract.md b/docs/features/doc-extract.md new file mode 100644 index 0000000..d1ecf18 --- /dev/null +++ b/docs/features/doc-extract.md @@ -0,0 +1,331 @@ +# 文档解析与本地 OCR + +## 1. 目标 + +GoodBuddy 需要用同一条可信文档解析链路服务以下场景： + +- 聊天附件问答； +- 知识库导入、同步、分块与来源定位； +- 后续的合同审阅、表格分析、演示文稿理解和文档转换。 + +文档解析不是对话模型的附属功能。它是主进程管理的独立基础能力，设置入口为“设置中心 / 文档解析”。 + +## 2. 当前基线 + +原生解析器已经支持： + +- UTF-8 文本、代码、配置、HTML； +- 带文本层的 PDF； +- DOCX 正文； +- XLSX 工作表 XML 与共享字符串； +- PPTX 幻灯片文字。 + +现有局限： + +- 纯扫描 PDF 没有文本层时无法提取内容； +- DOC、XLS、PPT 等旧版二进制 Office 格式不支持； +- Office 解析主要提取文字，不能完整保留表格、公式、图表和版面； +- 聊天附件和知识库直接调用底层解析函数，缺少可配置的统一工作流； +- 没有本地 OCR 模型状态、真实解析测试和按场景策略。 + +## 3. 产品原则 + +### 3.1 双通道解析 + +PDF 不是所有文档唯一的中间格式。解析应同时保留： + +1. 原生语义通道：标题、段落、单元格、公式、备注和对象关系； +2. 渲染视觉通道：页码、版面、图表、图片和 OCR 结果。 + +两条通道合并为统一文档结构。转换为 PDF 用于补充视觉信息，不得覆盖更可靠的原生语义结果。 + +### 3.2 场景工作流 + +| 场景 | 默认预设 | 行为 | +| --- | --- | --- | +| 聊天附件 | 自动解析 | 优先快速提取，文本不足时按需 OCR，有界截断后加入当前请求 | +| 知识库导入 | 完整索引 | 完整解析、按页或工作表定位、按需 OCR、分块与索引 | +| 扫描文档 | OCR | 页面渲染、文字识别、置信度与定位保留 | +| 表格分析 | 语义优先 | 单元格和值优先，PDF 或图片补充图表与打印布局 | +| 高保真审阅 | 视觉增强 | 原生解析、页面渲染、OCR 或视觉理解合并 | + +### 3.3 本地优先 + +- 文本层和本地 OCR 均在设备上处理； +- 本地处理不因 Ask 或 Execute 模式改变； +- OCR 来源必须在“本地模型 / 远程服务”之间明确选择； +- 配置并保存远程服务即表示用户选择该处理路径，不再增加逐场景授权； +- API 密钥只能保存在主进程加密设置中； +- 测试文件不得自动进入聊天或知识库。 + +## 4. 设置设计 + +设置中心新增“文档解析”分类，结构如下： + +1. 分类页头：“测试解析”“保存设置”； +2. 运行状态：原生解析、文档转换、本地 OCR； +3. 使用场景：聊天附件、知识库导入； +4. 文档转换； +5. OCR 识别； +6. 高级解析设置； + +OCR 模型区沿用语音模型管理模式： + +- 应用不内置模型权重； +- 用户按需从 ModelScope 下载，下载完成后离线使用； +- 显示来源、语言、运行时、模型体积、安装与校验状态； +- 联网设备可导出已安装模型 ZIP，离线或内网设备可直接导入； +- 支持下载进度、取消、删除、ZIP 导入导出、打开模型仓库和受管目录； +- “打开 ModelScope”直接显示在 OCR 模型卡片右上角，不使用手动导入折叠区； +- 模型操作即时生效，解析策略仍通过分类页头的“保存设置”提交。 + +### 4.1 第一阶段字段 + +- 聊天附件预设：`auto`、`fast-text`、`high-fidelity`； +- 知识库预设：`complete-index`、`fast-index`、`high-fidelity`； +- PDF OCR 策略：`auto`、`always`、`disabled`； +- OCR 来源：第一阶段固定为 `local`，远程服务入口禁用； +- 本地 OCR 模型：`pp-ocrv6-tiny`、`pp-ocrv6-small`、`pp-ocrv6-medium`； +- 单文档最大页数； +- OCR 并发数； +- 单页超时。 + +OCR 来源使用互斥选择。本地模型选中后才显示模型下拉列表、按需下载、导入和本地 OCR 参数；远程服务计划接入 MinerU、PaddleOCR-VL 等接口，第一阶段保持可读但禁用。来源选择本身就是用户的明确决策，不再显示额外的“隐私与云端处理”授权区。 + +## 5. 架构 + +```text +聊天附件 ─┐ + ├─ DocumentParsingService +知识库导入 ┘ ├─ NativeDocumentParser + ├─ PdfTextQualityEvaluator + ├─ PdfPageRenderer + ├─ LocalOcrProvider + ├─ DocumentConversionProvider + └─ ParsedDocument merger +``` + +`DocumentParsingService` 是唯一场景入口： + +```ts +type DocumentParsingPurpose = 'chat-attachment' | 'knowledge-index' + +type DocumentParsingService = { + parse( + name: string, + bytes: Buffer, + purpose: DocumentParsingPurpose, + signal?: AbortSignal + ): Promise +} +``` + +聊天上下文管理器与知识库服务依赖该接口，不直接选择 OCR Provider。 + +## 6. 统一结果 + +第一阶段兼容现有 `ParsedDocument`，并逐步扩展： + +```ts +type ParsedDocument = { + title: string + sourceFormat: string + content: string + sections: Array<{ + locator: string + content: string + method?: 'native' | 'ocr' | 'converted' | 'vision' + confidence?: number + }> + warnings?: string[] +} +``` + +定位字段必须对使用者有意义： + +- PDF：`第 3 页`； +- XLSX：`工作表：预算 / A1:F28`； +- PPTX：`幻灯片 5`； +- DOCX：标题路径或页码； +- 文本：`全文`。 + +## 7. 本地 OCR 基线 + +### 7.1 模型与运行时 + +全平台功能基线： + +- 模型：PP-OCRv6 ONNX/ORT； +- 轻量下载档位：Tiny，约 6 MiB，用于低资源设备和六平台离线链路； +- 推荐下载档位：Small，约 30 MiB，官方支持 50 种语言； +- 高精度下载档位：Medium，约 132 MiB，官方支持 50 种语言，但识别较慢且需要更多内存； +- 运行时：ONNX Runtime WebAssembly； +- 处理环境：隔离 Worker； +- 加速：WebGPU 或平台原生执行 Provider，仅作为可选层； +- 回退：任何加速失败后使用 WASM CPU。 + +需要覆盖的发布矩阵： + +- Windows x64、Windows arm64； +- macOS x64、macOS arm64； +- Linux x64、Linux arm64。 + +模型清单必须固定以下信息： + +- 上游仓库和不可变 revision； +- 文件名、字节数和 SHA-256； +- 模型族、语言、质量和速度； +- 许可证名称、完整许可证和来源； +- 检测模型、识别模型、字符字典的匹配关系。 + +运行时不得从 `main`、`latest` 或其他可变地址加载模型。 + +### 7.2 下载与安装 + +Tiny、Small 和 Medium 模型均由 PaddlePaddle 官方 ModelScope 仓库提供。Small 是默认推荐档位；Medium 面向更高识别质量，但具有更高内存占用和延迟。每个档位的检测模型、识别模型与字符字典配置分别使用固定提交，并在应用内记录文件字节数和 SHA-256。 + +下载流程： + +1. 主进程从固定 ModelScope `resolve//...` 地址读取文件； +2. 禁用凭据与缓存，限制重定向次数和单文件大小； +3. 写入受管目录下的随机临时安装目录； +4. 边下载边计算 SHA-256，并核对完整字节数； +5. 三个文件全部通过校验后写入安装清单； +6. 原子重命名为正式模型目录； +7. 失败、取消或退出时删除临时文件。 + +模型只在下载或用户显式打开仓库时访问网络。OCR 推理从受管目录读取已校验文件，不发起网络请求。 + +### 7.3 离线 ZIP 迁移 + +语音模型和 OCR 模型使用同一种离线迁移流程： + +1. 联网设备完成受信任来源下载和校验； +2. 在模型卡片选择“导出 ZIP”； +3. 将 ZIP 通过组织批准的介质传输到离线或内网设备； +4. 在相同模型的卡片选择“导入 ZIP”； +5. 主进程按当前应用内置目录重新校验，并在全部通过后原子安装。 + +ZIP 根目录包含模型文件和 `goodbuddy-model.json`。清单格式为 `goodbuddy-model-archive`，当前版本为 `1`，记录： + +- 模型类型：`speech` 或 `document-ocr`； +- 内置模型 ID 和显示名称； +- 文件名、角色、原始字节数和 SHA-256； +- 导出时间。 + +导出不能直接信任已有安装清单，必须重新读取并校验每个文件。导入不能只信任 ZIP 自声明内容，模型 ID、文件角色、字节数和哈希必须再次与当前应用内置目录完全匹配。导入通过后复用普通本地安装的受控临时目录和原子重命名路径。 + +归档处理使用有界流式读写，不把大型模型或整个展开结果复制到内存。主进程限制压缩包大小、条目数、清单大小、单文件大小和总展开大小，并拒绝： + +- 绝对路径、`..`、目录或嵌套路径； +- 大小写不敏感的重复条目； +- 未声明、缺失或角色不匹配的文件； +- 模型类型或模型 ID 不匹配； +- 解压后大小或 SHA-256 不匹配； +- 超过边界的压缩包和压缩炸弹。 + +取消文件对话框不会改变安装状态。导入和导出也不会切换当前语音/OCR 模型，不会隐式保存文档解析设置。 + +### 7.4 PDF 流程 + +1. 使用 PDF.js 读取每页文本层； +2. 评估有效字符数、乱码率和图片占比； +3. `auto` 模式只渲染文本不足的页面； +4. `always` 模式渲染所有页面； +5. Worker 将页面限制在配置的最大边长内； +6. OCR 返回文字、坐标和置信度； +7. 按页合并原生文本与 OCR，不重复可靠文本； +8. 达到页数、超时、取消或输出限制时停止并返回明确错误。 + +受密码保护、损坏或超限的 PDF 不得进入 OCR。 + +## 8. Office 与转换 + +### 8.1 新格式 + +- DOCX：正文、标题、表格、批注和图片关系； +- XLSX：工作表、单元格地址、值、公式、合并关系和图表； +- PPTX：幻灯片、文字对象、备注、图片和阅读顺序。 + +Office 内嵌图片 OCR 属于增强流程，不能替代原生结构解析。 + +### 8.2 旧格式 + +DOC、XLS、PPT 通过 `DocumentConversionProvider` 转换： + +1. 转换为 DOCX、XLSX 或 PPTX，供语义解析； +2. 转换为 PDF，供页码、版面和视觉解析； +3. 合并结果并记录转换警告。 + +本地 LibreOffice Provider 必须： + +- 在隔离子进程中运行； +- 禁用宏和网络； +- 使用单任务临时目录； +- 限制输入大小、输出大小、内存和超时； +- 在成功、失败、取消和退出时清理； +- 不接受用户提供的任意命令参数。 + +## 9. 安全边界 + +- 文件路径解析、读取、大小检查和格式校验在主进程完成； +- OCR Worker 只接收当前任务所需的有界页面图像和只读模型； +- 不向 Worker 暴露文件系统、Electron API、凭据或任意网络访问； +- 文档内容视为不可信数据，不解释其中的提示词为系统指令； +- 模型和转换程序必须固定版本并校验哈希； +- OCR 输出受字符数限制，错误不得包含绝对路径或未脱敏文档内容； +- 取消、超时和应用关闭必须终止待处理页面并释放模型会话。 + +## 10. 错误与回退 + +必须区分： + +- 不支持的格式； +- 文档损坏或受密码保护； +- 文本层为空但 OCR 未启用； +- OCR 模型不可用； +- OCR 超时或取消； +- 文档页数、大小或输出超限； +- 本地转换服务未配置； +- 所选远程 OCR 服务不可用或配置不完整。 + +`auto` 工作流可以从 OCR 回退到可靠的原生文本，但不能把空结果标记为成功。知识库导入失败时保留来源和可重试上下文。 + +## 11. 实施阶段 + +### 阶段一 + +- 新增文档解析设置分类和持久化契约； +- 建立 `DocumentParsingService`，供聊天和知识库共用； +- 将无文本 PDF 识别为可触发 OCR 的明确状态； +- 接入 PP-OCRv6 Tiny、Small、Medium 的 ModelScope 下载、校验、ZIP 离线迁移、删除与 WASM Worker； +- 实现真实文件测试和六平台验证入口。 + +### 阶段二 + +- 增强 DOCX、XLSX、PPTX 语义结构； +- 实现按页混合文本层与 OCR； +- 增加版面、表格和阅读顺序。 + +### 阶段三 + +- 增加 LibreOffice 和 API 转换 Provider； +- 支持 DOC、XLS、PPT； +- 增加 MinerU、PaddleOCR-VL 等远程 OCR 服务连接配置； +- 增加高保真工作流和解析结果预览。 + +## 12. 验收 + +- 同一份扫描 PDF 可从聊天附件和知识库得到一致的逐页文本； +- 文本型 PDF 在 `auto` 模式下不运行 OCR； +- 本地 OCR 在六个平台和两种架构上完全离线运行； +- 模型文件损坏时拒绝加载并显示可恢复错误； +- 未安装模型时扫描文档提示用户前往“文档解析”下载，文本型文档仍可原生解析； +- 下载中可显示文件与总进度并允许取消，失败或取消后不留下已安装状态； +- ModelScope 下载与 ZIP 导入均经过同一大小和 SHA-256 校验； +- 语音和 OCR 模型可在联网设备导出 ZIP，并在离线设备导入后完成真实推理； +- 路径穿越、未知条目、错误模型 ID、篡改文件和超限 ZIP 均被拒绝； +- 超页数、超时、取消和关闭不会留下运行任务； +- 测试解析不会创建聊天消息或知识库文档； +- 选择本地模型时没有任何文档上传； +- 文档中的提示词不会改变系统、模式或工具权限。 diff --git a/electron.vite.config.ts b/electron.vite.config.ts index 12a8658..644d274 100644 --- a/electron.vite.config.ts +++ b/electron.vite.config.ts @@ -34,6 +34,9 @@ export default defineConfig({ '@shared': resolve('src/shared') } }, + worker: { + format: 'es' + }, plugins: [react()] } }) diff --git a/package-lock.json b/package-lock.json index 1c334fe..27a5c44 100644 --- a/package-lock.json +++ b/package-lock.json @@ -20,7 +20,9 @@ "i18next": "^25.10.10", "json5": "^2.2.3", "lucide-react": "^1.27.0", + "onnxruntime-web": "^1.23.2", "pdfjs-dist": "^6.2.108", + "ppu-paddle-ocr": "^6.4.0", "qrcode": "^1.5.4", "quill": "^2.0.3", "react": "^19.2.8", @@ -2325,7 +2327,6 @@ "resolved": "https://registry.npmjs.org/@napi-rs/canvas/-/canvas-1.0.3.tgz", "integrity": "sha512-OlI657a5XXvKGFX7kNeIzJ8rO7IXt87Mqu2H8rXE46viAuOfum/JA7ysX7+eBhxNKznT+RCZh418mndlcFX3+w==", "license": "MIT", - "optional": true, "workspaces": [ "e2e/*" ], @@ -2659,6 +2660,63 @@ "node": ">=14.18.0" } }, + "node_modules/@protobufjs/aspromise": { + "version": "1.1.2", + "resolved": "https://registry.npmjs.org/@protobufjs/aspromise/-/aspromise-1.1.2.tgz", + "integrity": "sha512-j+gKExEuLmKwvz3OgROXtrJ2UG2x8Ch2YZUxahh+s1F2HZ+wAceUNLkvy6zKCPVRkU++ZWQrdxsUeQXmcg4uoQ==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/base64": { + "version": "1.1.2", + "resolved": "https://registry.npmjs.org/@protobufjs/base64/-/base64-1.1.2.tgz", + "integrity": "sha512-AZkcAA5vnN/v4PDqKyMR5lx7hZttPDgClv83E//FMNhR2TMcLUhfRUBHCmSl0oi9zMgDDqRUJkSxO3wm85+XLg==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/codegen": { + "version": "2.0.5", + "resolved": "https://registry.npmjs.org/@protobufjs/codegen/-/codegen-2.0.5.tgz", + "integrity": "sha512-zgXFLzW3Ap33e6d0Wlj4MGIm6Ce8O89n/apUaGNB/jx+hw+ruWEp7EwGUshdLKVRCxZW12fp9r40E1mQrf/34g==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/eventemitter": { + "version": "1.1.1", + "resolved": "https://registry.npmjs.org/@protobufjs/eventemitter/-/eventemitter-1.1.1.tgz", + "integrity": "sha512-vW1GmwMZNnL+gMRaovlh9yZX74kc+TTU3FObkkurpMaRtBfLP3ldjS9KQWlwZgraRE0+dheEEoAxdzcJQ8eXZg==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/fetch": { + "version": "1.1.1", + "resolved": "https://registry.npmjs.org/@protobufjs/fetch/-/fetch-1.1.1.tgz", + "integrity": "sha512-GpptLrs57adMSuHi3VNj0mAF8dwh36LMaYF6XyJ6JMWlVsc+t42tm1HSEDmOs3A8fC9yyeisgLhsTVQokOZ0zw==", + "license": "BSD-3-Clause", + "dependencies": { + "@protobufjs/aspromise": "^1.1.1" + } + }, + "node_modules/@protobufjs/float": { + "version": "1.0.2", + "resolved": "https://registry.npmjs.org/@protobufjs/float/-/float-1.0.2.tgz", + "integrity": "sha512-Ddb+kVXlXst9d+R9PfTIxh1EdNkgoRe5tOX6t01f1lYWOvJnSPDBlG241QLzcyPdoNTsblLUdujGSE4RzrTZGQ==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/path": { + "version": "1.1.2", + "resolved": "https://registry.npmjs.org/@protobufjs/path/-/path-1.1.2.tgz", + "integrity": "sha512-6JOcJ5Tm08dOHAbdR3GrvP+yUUfkjG5ePsHYczMFLq3ZmMkAD98cDgcT2iA1lJ9NVwFd4tH/iSSoe44YWkltEA==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/pool": { + "version": "1.1.0", + "resolved": "https://registry.npmjs.org/@protobufjs/pool/-/pool-1.1.0.tgz", + "integrity": "sha512-0kELaGSIDBKvcgS4zkjz1PeddatrjYcmMWOlAuAPwAeccUrPHdUqo/J6LiymHHEiJT5NrF1UVwxY14f+fy4WQw==", + "license": "BSD-3-Clause" + }, + "node_modules/@protobufjs/utf8": { + "version": "1.1.2", + "resolved": "https://registry.npmjs.org/@protobufjs/utf8/-/utf8-1.1.2.tgz", + "integrity": "sha512-b1UQwcEZ4yCnMCD8DAL1VlbvBJE9/IX4FTIp7BG1xYpf29SLazLSrqUkj4w7Y5y7cCVP6E5tcqqcI0xemPkHug==", + "license": "BSD-3-Clause" + }, "node_modules/@rolldown/pluginutils": { "version": "1.0.0-rc.3", "resolved": "https://registry.npmjs.org/@rolldown/pluginutils/-/pluginutils-1.0.0-rc.3.tgz", @@ -3104,6 +3162,12 @@ "node": ">=10" } }, + "node_modules/@techstark/opencv-js": { + "version": "5.0.0-release.1", + "resolved": "https://registry.npmjs.org/@techstark/opencv-js/-/opencv-js-5.0.0-release.1.tgz", + "integrity": "sha512-PIm+eB0MFtieXoNC2GRao0dv/02sehG+Nv2nSW5D6pQm6J/4WqvDHm0RyoqOmGYQm67jdGiaOdIeTShCY3PIUg==", + "license": "Apache-2.0" + }, "node_modules/@testing-library/dom": { "version": "10.4.1", "resolved": "https://registry.npmjs.org/@testing-library/dom/-/dom-10.4.1.tgz", @@ -3512,7 +3576,6 @@ "version": "26.1.2", "resolved": "https://registry.npmjs.org/@types/node/-/node-26.1.2.tgz", "integrity": "sha512-Vu4a5UFA9rIIFJ7rB/Vaafh9lrCQszopTCx6KjFboXTGQbPNasehVR5TEiithSDGyd1DEiUByggTZsg8jukeIg==", - "dev": true, "license": "MIT", "dependencies": { "undici-types": "~8.3.0" @@ -6794,6 +6857,12 @@ "node": ">=16" } }, + "node_modules/flatbuffers": { + "version": "25.9.23", + "resolved": "https://registry.npmjs.org/flatbuffers/-/flatbuffers-25.9.23.tgz", + "integrity": "sha512-MI1qs7Lo4Syw0EOzUl0xjs2lsoeqFku44KpngfIduHBYvzm8h2+7K8YMQh1JtVVVrUvhLpNwqVi4DERegUJhPQ==", + "license": "Apache-2.0" + }, "node_modules/flatted": { "version": "3.4.3", "resolved": "https://registry.npmjs.org/flatted/-/flatted-3.4.3.tgz", @@ -7170,6 +7239,12 @@ "lodash": "^4.17.15" } }, + "node_modules/guid-typescript": { + "version": "1.0.9", + "resolved": "https://registry.npmjs.org/guid-typescript/-/guid-typescript-1.0.9.tgz", + "integrity": "sha512-Y8T4vYhEfwJOTbouREvG+3XDsjr8E3kIr7uf+JZ0BYloFsttiHU0WfvANVsR7TxNUJa/WpCnw/Ino/p+DeBhBQ==", + "license": "ISC" + }, "node_modules/has-flag": { "version": "4.0.0", "resolved": "https://registry.npmjs.org/has-flag/-/has-flag-4.0.0.tgz", @@ -8044,6 +8119,12 @@ "deprecated": "This package is deprecated. Use require('node:util').isDeepStrictEqual instead.", "license": "MIT" }, + "node_modules/long": { + "version": "5.3.2", + "resolved": "https://registry.npmjs.org/long/-/long-5.3.2.tgz", + "integrity": "sha512-mNAgZ1GmyNhD7AuqnTG3/VQ26o760+ZYBPKjPvugO8+nLbYfX6TVpJPseBvopbdY+qpZ/lKUnmEc1LeZYS3QAA==", + "license": "Apache-2.0" + }, "node_modules/longest-streak": { "version": "3.1.0", "resolved": "https://registry.npmjs.org/longest-streak/-/longest-streak-3.1.0.tgz", @@ -9457,6 +9538,26 @@ "wrappy": "1" } }, + "node_modules/onnxruntime-common": { + "version": "1.23.2", + "resolved": "https://registry.npmjs.org/onnxruntime-common/-/onnxruntime-common-1.23.2.tgz", + "integrity": "sha512-5LFsC9Dukzp2WV6kNHYLNzp8sT6V02IubLCbzw2Xd6X5GOlr65gAX6xiJwyi2URJol/s71gaQLC5F2C25AAR2w==", + "license": "MIT" + }, + "node_modules/onnxruntime-web": { + "version": "1.23.2", + "resolved": "https://registry.npmjs.org/onnxruntime-web/-/onnxruntime-web-1.23.2.tgz", + "integrity": "sha512-T09JUtMn+CZLk3mFwqiH0lgQf+4S7+oYHHtk6uhaYAAJI95bTcKi5bOOZYwORXfS/RLZCjDDEXGWIuOCAFlEjg==", + "license": "MIT", + "dependencies": { + "flatbuffers": "^25.1.24", + "guid-typescript": "^1.0.9", + "long": "^5.2.3", + "onnxruntime-common": "1.23.2", + "platform": "^1.3.6", + "protobufjs": "^7.2.4" + } + }, "node_modules/opencode-ai": { "version": "1.18.9", "resolved": "https://registry.npmjs.org/opencode-ai/-/opencode-ai-1.18.9.tgz", @@ -9932,6 +10033,12 @@ "url": "https://paulmillr.com/funding/" } }, + "node_modules/platform": { + "version": "1.3.6", + "resolved": "https://registry.npmjs.org/platform/-/platform-1.3.6.tgz", + "integrity": "sha512-fnWVljUchTro6RiCFvCXBbNhJc2NijN7oIQxbwsyL0buWJPG85v81ehlHI9fXrJsMNgTofEoWIQeClKpgxFLrg==", + "license": "MIT" + }, "node_modules/plist": { "version": "3.1.0", "resolved": "https://registry.npmjs.org/plist/-/plist-3.1.0.tgz", @@ -10051,6 +10158,56 @@ "node": "^12.20.0 || >=14" } }, + "node_modules/ppu-ocv": { + "version": "4.0.0", + "resolved": "https://registry.npmjs.org/ppu-ocv/-/ppu-ocv-4.0.0.tgz", + "integrity": "sha512-ol+S9/KLZ0aTV0xbC8+ZB0iM+FYux8ujoULeQoYY2IDBZj4LAuujJUkzHzUA2O3KzHs17KuuIzV/w4/htPdn3w==", + "license": "MIT", + "dependencies": { + "@napi-rs/canvas": "^1.0.0", + "@techstark/opencv-js": "^5.0.0-release.1" + }, + "peerDependencies": { + "@shopify/react-native-skia": ">=1.0.0" + }, + "peerDependenciesMeta": { + "@shopify/react-native-skia": { + "optional": true + } + } + }, + "node_modules/ppu-paddle-ocr": { + "version": "6.4.0", + "resolved": "https://registry.npmjs.org/ppu-paddle-ocr/-/ppu-paddle-ocr-6.4.0.tgz", + "integrity": "sha512-Llhlh6zIDbbvgG+6zVKLaoRdi8sQzH8hkh6YB6WewFjK7jhbNKfX8Rel2w2g7EKELhLhgg2zCTaZFM9loeWLVQ==", + "license": "MIT", + "dependencies": { + "ppu-ocv": "^4.0.0" + }, + "bin": { + "ppu-paddle-ocr": "cli/index.js" + }, + "peerDependencies": { + "@shopify/react-native-skia": ">=1.0.0", + "onnxruntime-node": "^1.23.2", + "onnxruntime-react-native": "^1.23.2", + "onnxruntime-web": "^1.23.2" + }, + "peerDependenciesMeta": { + "@shopify/react-native-skia": { + "optional": true + }, + "onnxruntime-node": { + "optional": true + }, + "onnxruntime-react-native": { + "optional": true + }, + "onnxruntime-web": { + "optional": true + } + } + }, "node_modules/prelude-ls": { "version": "1.2.1", "resolved": "https://registry.npmjs.org/prelude-ls/-/prelude-ls-1.2.1.tgz", @@ -10154,6 +10311,29 @@ "url": "https://github.com/sponsors/wooorm" } }, + "node_modules/protobufjs": { + "version": "7.6.5", + "resolved": "https://registry.npmjs.org/protobufjs/-/protobufjs-7.6.5.tgz", + "integrity": "sha512-/FPD0nUc9jH6rfFjji9IBqOz4pcSE3CsT1m7Ep6Mdb0LxSUMj8hgl6GomOvZzpNpAqqGaXA0P3VSrZLFzIhQrw==", + "hasInstallScript": true, + "license": "BSD-3-Clause", + "dependencies": { + "@protobufjs/aspromise": "^1.1.2", + "@protobufjs/base64": "^1.1.2", + "@protobufjs/codegen": "^2.0.5", + "@protobufjs/eventemitter": "^1.1.1", + "@protobufjs/fetch": "^1.1.1", + "@protobufjs/float": "^1.0.2", + "@protobufjs/path": "^1.1.2", + "@protobufjs/pool": "^1.1.0", + "@protobufjs/utf8": "^1.1.1", + "@types/node": ">=13.7.0", + "long": "^5.3.2" + }, + "engines": { + "node": ">=12.0.0" + } + }, "node_modules/proxy-addr": { "version": "2.0.7", "resolved": "https://registry.npmjs.org/proxy-addr/-/proxy-addr-2.0.7.tgz", @@ -11737,7 +11917,6 @@ "version": "8.3.0", "resolved": "https://registry.npmjs.org/undici-types/-/undici-types-8.3.0.tgz", "integrity": "sha512-j375ScV60dom+YkPFIfTLcOiPxkN/buHz5GobjLhixFuANaNs3C9l4GmrWqejgXWJ7BbJcFYpTEUkS1Ge8bpZQ==", - "dev": true, "license": "MIT" }, "node_modules/unicorn-magic": { diff --git a/package.json b/package.json index 37df79b..ad72082 100644 --- a/package.json +++ b/package.json @@ -98,6 +98,18 @@ { "from": "node_modules/@fontsource-variable/noto-sans-sc/LICENSE", "to": "licenses/noto-sans-sc-OFL-1.1.txt" + }, + { + "from": "node_modules/ppu-paddle-ocr/LICENSE", + "to": "licenses/ppu-paddle-ocr-MIT.txt" + }, + { + "from": "node_modules/ppu-ocv/LICENSE", + "to": "licenses/ppu-ocv-MIT.txt" + }, + { + "from": "node_modules/onnxruntime-web/LICENSE", + "to": "licenses/onnxruntime-web-MIT.txt" } ], "win": { @@ -148,7 +160,9 @@ "i18next": "^25.10.10", "json5": "^2.2.3", "lucide-react": "^1.27.0", + "onnxruntime-web": "^1.23.2", "pdfjs-dist": "^6.2.108", + "ppu-paddle-ocr": "^6.4.0", "qrcode": "^1.5.4", "quill": "^2.0.3", "react": "^19.2.8", diff --git a/src/main/context-manager.ts b/src/main/context-manager.ts index 05dd197..2e4bdaf 100644 --- a/src/main/context-manager.ts +++ b/src/main/context-manager.ts @@ -24,6 +24,7 @@ import type { } from './agent/runtime' import { encodeBoundedJpeg } from './bounded-jpeg' import { parseDocument } from './knowledge/document-parser' +import type { ParsedDocument } from './knowledge/document-parser' type StoredTextContext = ContextAttachment & { kind: 'text' @@ -94,7 +95,7 @@ function truncateUtf8(value: string, maximumBytes: number): string { } function formatParsedDocument( - sections: Awaited>['sections'] + sections: ParsedDocument['sections'] ): string { return sections .map( @@ -121,6 +122,23 @@ function remoteAttachmentName(value: string): string { export class ContextManager { private readonly contexts = new Map() private totalBytes = 0 + private readonly documentParser: ( + name: string, + buffer: Buffer, + purpose: 'chat-attachment' + ) => Promise + + constructor(options?: { + parseDocument?: ( + name: string, + buffer: Buffer, + purpose: 'chat-attachment' + ) => Promise + }) { + this.documentParser = + options?.parseDocument ?? + ((name, buffer) => parseDocument(name, buffer)) + } private toPublic(context: StoredContext): ContextAttachment { return { @@ -245,7 +263,11 @@ export class ContextManager { ) } if (supportedDocumentExtensions.has(extension)) { - const parsed = await parseDocument(name, data) + const parsed = await this.documentParser( + name, + data, + 'chat-attachment' + ) return this.storeText( name, truncateUtf8( @@ -340,9 +362,10 @@ export class ContextManager { ) { throw new Error('PDF 或 Office 文档必须小于 20MB 且不能是目录') } - const parsed = await parseDocument( + const parsed = await this.documentParser( basename(canonicalPath), - await handle.readFile() + await handle.readFile(), + 'chat-attachment' ) attachments.push( this.storeText( diff --git a/src/main/document-ocr-broker.test.ts b/src/main/document-ocr-broker.test.ts new file mode 100644 index 0000000..0394e9a --- /dev/null +++ b/src/main/document-ocr-broker.test.ts @@ -0,0 +1,53 @@ +import { describe, expect, it, vi } from 'vitest' +import { ipcChannels } from '../shared/ipc-channels' +import { DocumentOcrBroker } from './document-ocr-broker' + +function request() { + return { + modelId: 'pp-ocrv6-tiny', + fileName: 'scan.pdf', + mimeType: 'application/pdf' as const, + data: new ArrayBuffer(8), + maximumPages: 10, + pageNumbers: [1], + pageTimeoutSeconds: 60 + } +} + +describe('DocumentOcrBroker', () => { + it('forwards an AbortSignal cancellation to the renderer', async () => { + const send = vi.fn() + const broker = new DocumentOcrBroker({ + isDestroyed: vi.fn(() => false), + webContents: { send } + } as never) + const controller = new AbortController() + const result = broker.recognize(request(), controller.signal) + const ocrRequest = send.mock.calls.find( + ([channel]) => channel === ipcChannels.documentParsingOcrRequest + )?.[1] as { requestId: string } + + controller.abort() + + await expect(result).rejects.toThrow('OCR 解析已取消') + expect(send).toHaveBeenCalledWith( + ipcChannels.documentParsingOcrCancel, + ocrRequest.requestId + ) + broker.dispose() + }) + + it('rejects a request that is already cancelled', () => { + const broker = new DocumentOcrBroker({ + isDestroyed: vi.fn(() => false), + webContents: { send: vi.fn() } + } as never) + const controller = new AbortController() + controller.abort() + + expect(() => + broker.recognize(request(), controller.signal) + ).toThrow('OCR 解析已取消') + broker.dispose() + }) +}) diff --git a/src/main/document-ocr-broker.ts b/src/main/document-ocr-broker.ts new file mode 100644 index 0000000..5fd6664 --- /dev/null +++ b/src/main/document-ocr-broker.ts @@ -0,0 +1,131 @@ +import type { BrowserWindow } from 'electron' +import { ipcChannels } from '../shared/ipc-channels' +import { + documentOcrFailureSchema, + documentOcrRequestSchema, + documentOcrResultSchema, + type DocumentOcrRequest, + type DocumentOcrResult +} from '../shared/document-parsing-contracts' + +type PendingRequest = { + resolve: (result: DocumentOcrResult) => void + reject: (error: Error) => void + timer: ReturnType + detachAbort: () => void +} + +const maximumPendingRequests = 4 +const maximumTotalTimeoutMs = 10 * 60 * 1_000 + +export class DocumentOcrBroker { + private readonly pending = new Map() + private disposed = false + + constructor(private readonly window: BrowserWindow) {} + + recognize( + input: Omit, + signal?: AbortSignal + ): Promise { + if (this.disposed || this.window.isDestroyed()) { + throw new Error('OCR 渲染服务不可用') + } + if (this.pending.size >= maximumPendingRequests) { + throw new Error('OCR 任务过多，请稍后重试') + } + const request = documentOcrRequestSchema.parse({ + ...input, + requestId: crypto.randomUUID() + }) + if (signal?.aborted) { + throw new Error('OCR 解析已取消') + } + const timeoutMs = Math.min( + maximumTotalTimeoutMs, + Math.max( + request.pageTimeoutSeconds * 1_000, + request.pageTimeoutSeconds * + request.maximumPages * + 1_000 + ) + ) + return new Promise((resolve, reject) => { + const cancel = (message: string): void => { + const pending = this.pending.get(request.requestId) + if (!pending) { + return + } + clearTimeout(pending.timer) + pending.detachAbort() + this.pending.delete(request.requestId) + this.window.webContents.send( + ipcChannels.documentParsingOcrCancel, + request.requestId + ) + reject(new Error(message)) + } + const timer = setTimeout(() => { + cancel('OCR 解析超时') + }, timeoutMs) + const onAbort = (): void => cancel('OCR 解析已取消') + signal?.addEventListener('abort', onAbort, { once: true }) + this.pending.set(request.requestId, { + resolve, + reject, + timer, + detachAbort: () => + signal?.removeEventListener('abort', onAbort) + }) + if (signal?.aborted) { + cancel('OCR 解析已取消') + return + } + this.window.webContents.send( + ipcChannels.documentParsingOcrRequest, + request + ) + }) + } + + respond(input: unknown): void { + const result = documentOcrResultSchema.safeParse(input) + const failure = result.success + ? undefined + : documentOcrFailureSchema.safeParse(input) + const requestId = result.success + ? result.data.requestId + : failure?.success + ? failure.data.requestId + : undefined + if (!requestId) { + throw new Error('OCR 响应无效') + } + const pending = this.pending.get(requestId) + if (!pending) { + return + } + clearTimeout(pending.timer) + pending.detachAbort() + this.pending.delete(requestId) + if (result.success) { + pending.resolve(result.data) + } else { + if (!failure?.success) { + pending.reject(new Error('OCR 响应无效')) + return + } + pending.reject(new Error(failure.data.error)) + } + } + + dispose(): void { + this.disposed = true + for (const pending of this.pending.values()) { + clearTimeout(pending.timer) + pending.detachAbort() + pending.reject(new Error('OCR 解析已取消')) + } + this.pending.clear() + } +} diff --git a/src/main/document-ocr-model-catalog.ts b/src/main/document-ocr-model-catalog.ts new file mode 100644 index 0000000..51198c8 --- /dev/null +++ b/src/main/document-ocr-model-catalog.ts @@ -0,0 +1,204 @@ +import { + documentOcrModelCatalogEntrySchema, + type DocumentOcrModelCatalogEntry +} from '../shared/document-parsing-contracts' + +const detectionRevision = + '7d7f5d128d9309ebf6de4f21f404dd583afdbae3' +const recognitionRevision = + 'afba04b618200c5f4824531c6e42c957c6439d9a' +const smallDetectionRevision = + '956a0b620a4017cc04056c692be1703b0025d028' +const smallRecognitionRevision = + '296d43bc0ebced0fd9c605174aa5962e49810ab6' +const mediumDetectionRevision = + 'c317b40325be40bfaaff58c8dcece2a075294f8a' +const mediumRecognitionRevision = + 'db5d610d492a14e3c34dc1fd4e9339bd369f79e6' + +export const DOCUMENT_OCR_MODEL_CATALOG: readonly DocumentOcrModelCatalogEntry[] = + documentOcrModelCatalogEntrySchema.array().parse([ + { + id: 'pp-ocrv6-tiny', + displayName: 'PP-OCRv6 Tiny', + description: + 'PaddleOCR 官方轻量中文 OCR 模型，适合扫描 PDF 和图片的本地 CPU 识别。', + languages: ['中文', '英语'], + runtime: 'onnxruntime-web-wasm', + quality: 'basic', + speed: 'fast', + recommended: false, + repositoryUrl: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_tiny_rec_onnx', + license: { + name: 'Apache License 2.0', + notice: + '检测与识别模型由 PaddlePaddle 在 ModelScope 发布，使用前请阅读模型仓库及 PaddleOCR 的许可证说明。', + url: 'https://github.com/PaddlePaddle/PaddleOCR/blob/main/LICENSE' + }, + files: [ + { + name: 'detection.onnx', + role: 'detection', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_tiny_det_onnx/resolve/' + + `${detectionRevision}/inference.onnx`, + size: 1_780_590, + sha256: + '193bab7a04fca699a6c82e6abb5b81bdb28177f0abd4062552b04908dafb19f8' + } + }, + { + name: 'recognition.onnx', + role: 'recognition', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_tiny_rec_onnx/resolve/' + + `${recognitionRevision}/inference.onnx`, + size: 4_462_639, + sha256: + '9ef676d6ed3c88256a2d92c640c44f25b0c40947e111b14b8be8f594091563e6' + } + }, + { + name: 'dictionary.yml', + role: 'dictionary', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_tiny_rec_onnx/resolve/' + + `${recognitionRevision}/inference.yml`, + size: 55_571, + sha256: + '66170210bad538e83fff3c4a3867e547d6bf20b50d64b20347c4b913f3034ea1' + } + } + ] + }, + { + id: 'pp-ocrv6-small', + displayName: 'PP-OCRv6 Small', + description: + 'PaddleOCR 官方 50 语言 OCR 模型，在识别质量、速度和本地资源占用之间取得平衡。', + languages: ['50 种语言'], + runtime: 'onnxruntime-web-wasm', + quality: 'balanced', + speed: 'balanced', + recommended: true, + repositoryUrl: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_small_rec_onnx', + license: { + name: 'Apache License 2.0', + notice: + '检测与识别模型由 PaddlePaddle 在 ModelScope 发布，使用前请阅读模型仓库及 PaddleOCR 的许可证说明。', + url: 'https://github.com/PaddlePaddle/PaddleOCR/blob/main/LICENSE' + }, + files: [ + { + name: 'detection.onnx', + role: 'detection', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_small_det_onnx/resolve/' + + `${smallDetectionRevision}/inference.onnx`, + size: 9_880_512, + sha256: + 'd73e0058b7a8086bbd57f3d10b8bcd4ff95363f67e06e2762b5e814fe9c9410e' + } + }, + { + name: 'recognition.onnx', + role: 'recognition', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_small_rec_onnx/resolve/' + + `${smallRecognitionRevision}/inference.onnx`, + size: 21_159_378, + sha256: + '5435fd747c9e0efe15a96d0b378d5bd157e9492ed8fd80edf08f30d02fa24634' + } + }, + { + name: 'dictionary.yml', + role: 'dictionary', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_small_rec_onnx/resolve/' + + `${smallRecognitionRevision}/inference.yml`, + size: 150_579, + sha256: + 'ab078671bb49f06228eadccd34f1bb501e157f7a047095ffb943ba81512c77d1' + } + } + ] + }, + { + id: 'pp-ocrv6-medium', + displayName: 'PP-OCRv6 Medium', + description: + 'PaddleOCR 官方 50 语言高质量 OCR 模型，识别较慢，并需要更多内存且具有更高延迟。', + languages: ['50 种语言'], + runtime: 'onnxruntime-web-wasm', + quality: 'high', + speed: 'slow', + recommended: false, + repositoryUrl: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_medium_rec_onnx', + license: { + name: 'Apache License 2.0', + notice: + '检测与识别模型由 PaddlePaddle 在 ModelScope 发布，使用前请阅读模型仓库及 PaddleOCR 的许可证说明。', + url: 'https://github.com/PaddlePaddle/PaddleOCR/blob/main/LICENSE' + }, + files: [ + { + name: 'detection.onnx', + role: 'detection', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_medium_det_onnx/resolve/' + + `${mediumDetectionRevision}/inference.onnx`, + size: 62_032_837, + sha256: + 'eb13b44b25bb36f89528b68720af8a61d9cf381176107f465db1757b65d086e1' + } + }, + { + name: 'recognition.onnx', + role: 'recognition', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_medium_rec_onnx/resolve/' + + `${mediumRecognitionRevision}/inference.onnx`, + size: 76_554_979, + sha256: + '9c09abf0957f7968c7586464b7397b84ad2387a0497a351af40e9acc71b673ba' + } + }, + { + name: 'dictionary.yml', + role: 'dictionary', + download: { + url: + 'https://modelscope.cn/models/PaddlePaddle/' + + 'PP-OCRv6_medium_rec_onnx/resolve/' + + `${mediumRecognitionRevision}/inference.yml`, + size: 150_580, + sha256: + '991b700facf5b50a7de193468207d5f4255b538dde0d312ae3b7c7a9b6873129' + } + } + ] + } + ]) diff --git a/src/main/document-ocr-model-manager.test.ts b/src/main/document-ocr-model-manager.test.ts new file mode 100644 index 0000000..0a25a3b --- /dev/null +++ b/src/main/document-ocr-model-manager.test.ts @@ -0,0 +1,341 @@ +import { createHash } from 'node:crypto' +import { + mkdtemp, + mkdir, + rm, + writeFile +} from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { afterEach, describe, expect, it, vi } from 'vitest' +import type { DocumentOcrModelCatalogEntry } from '../shared/document-parsing-contracts' +import { DOCUMENT_OCR_MODEL_CATALOG } from './document-ocr-model-catalog' +import { + DocumentOcrModelManager, + extractPaddleCharacterDictionary +} from './document-ocr-model-manager' + +const temporaryDirectories: string[] = [] + +function sha256(value: Uint8Array): string { + return createHash('sha256').update(value).digest('hex') +} + +function dictionaryYaml(): Uint8Array { + const characters = [ + "'!'", + "'\"'", + "''''", + ...Array.from({ length: 120 }, (_, index) => + String.fromCodePoint(0x4e00 + index) + ) + ] + return Buffer.from( + `PostProcess:\n name: CTCLabelDecode\n character_dict:\n${characters + .map((character) => ` - ${character}`) + .join('\n')}\n`, + 'utf8' + ) +} + +function catalog( + detection: Uint8Array, + recognition: Uint8Array, + dictionary: Uint8Array +): readonly DocumentOcrModelCatalogEntry[] { + const files = [ + { + name: 'detection.onnx', + role: 'detection' as const, + bytes: detection + }, + { + name: 'recognition.onnx', + role: 'recognition' as const, + bytes: recognition + }, + { + name: 'dictionary.yml', + role: 'dictionary' as const, + bytes: dictionary + } + ] + return [ + { + id: 'pp-ocrv6-tiny', + displayName: 'PP-OCRv6 Tiny', + description: 'Test OCR model catalog entry.', + languages: ['中文', '英语'], + runtime: 'onnxruntime-web-wasm', + quality: 'balanced', + speed: 'fast', + recommended: true, + repositoryUrl: + 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_tiny_rec_onnx', + license: { + name: 'Apache License 2.0', + notice: 'Test license notice.', + url: 'https://example.com/license' + }, + files: files.map((file) => ({ + name: file.name, + role: file.role, + download: { + url: `https://modelscope.cn/models/example/resolve/revision/${file.name}`, + size: file.bytes.byteLength, + sha256: sha256(file.bytes) + } + })) + } + ] +} + +async function createManager( + bytes?: { + detection: Uint8Array + recognition: Uint8Array + dictionary: Uint8Array + } +): Promise<{ + directory: string + manager: DocumentOcrModelManager + modelBytes: { + detection: Uint8Array + recognition: Uint8Array + dictionary: Uint8Array + } +}> { + const directory = await mkdtemp( + join(tmpdir(), 'goodbuddy-document-ocr-model-') + ) + temporaryDirectories.push(directory) + const modelBytes = bytes ?? { + detection: Buffer.from('detection model'), + recognition: Buffer.from('recognition model'), + dictionary: dictionaryYaml() + } + const testCatalog = catalog( + modelBytes.detection, + modelBytes.recognition, + modelBytes.dictionary + ) + const entry = testCatalog[0] + if (!entry) { + throw new Error('Test OCR catalog is empty') + } + const files = new Map( + entry.files.map((file) => [ + file.download.url, + modelBytes[file.role] + ]) + ) + const transport = vi.fn(async (input: string | URL | Request) => { + const url = + input instanceof Request ? input.url : input.toString() + const body = files.get(url) + if (!body) { + return new Response(null, { status: 404 }) + } + return new Response(body, { + status: 200, + headers: { + 'content-length': String(body.byteLength) + } + }) + }) as unknown as typeof fetch + return { + directory, + manager: new DocumentOcrModelManager({ + userDataDirectory: directory, + fetch: transport, + catalog: testCatalog + }), + modelBytes + } +} + +afterEach(async () => { + await Promise.all( + temporaryDirectories.splice(0).map((directory) => + rm(directory, { recursive: true, force: true }) + ) + ) +}) + +describe('DocumentOcrModelManager', () => { + it('uses immutable SHA-256 verified ModelScope catalog files', () => { + expect(DOCUMENT_OCR_MODEL_CATALOG).toHaveLength(3) + expect( + new Set(DOCUMENT_OCR_MODEL_CATALOG.map((entry) => entry.id)).size + ).toBe(3) + expect( + DOCUMENT_OCR_MODEL_CATALOG.filter((entry) => entry.recommended).map( + (entry) => entry.id + ) + ).toEqual(['pp-ocrv6-small']) + + for (const entry of DOCUMENT_OCR_MODEL_CATALOG) { + for (const file of entry.files) { + expect(file.download.url).toMatch( + /^https:\/\/modelscope\.cn\/models\/PaddlePaddle\/[^/]+\/resolve\/[a-f0-9]{40}\/[^/]+$/u + ) + expect(file.download.sha256).toMatch(/^[a-f0-9]{64}$/u) + expect(file.download.size).toBeGreaterThan(0) + } + } + + expect( + DOCUMENT_OCR_MODEL_CATALOG.find( + (entry) => entry.id === 'pp-ocrv6-small' + ) + ).toMatchObject({ + languages: ['50 种语言'], + quality: 'balanced', + speed: 'balanced', + recommended: true, + files: [ + { + role: 'detection', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_small_det_onnx/resolve/956a0b620a4017cc04056c692be1703b0025d028/inference.onnx', + size: 9_880_512, + sha256: + 'd73e0058b7a8086bbd57f3d10b8bcd4ff95363f67e06e2762b5e814fe9c9410e' + } + }, + { + role: 'recognition', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_small_rec_onnx/resolve/296d43bc0ebced0fd9c605174aa5962e49810ab6/inference.onnx', + size: 21_159_378, + sha256: + '5435fd747c9e0efe15a96d0b378d5bd157e9492ed8fd80edf08f30d02fa24634' + } + }, + { + role: 'dictionary', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_small_rec_onnx/resolve/296d43bc0ebced0fd9c605174aa5962e49810ab6/inference.yml', + size: 150_579, + sha256: + 'ab078671bb49f06228eadccd34f1bb501e157f7a047095ffb943ba81512c77d1' + } + } + ] + }) + expect( + DOCUMENT_OCR_MODEL_CATALOG.find( + (entry) => entry.id === 'pp-ocrv6-medium' + ) + ).toMatchObject({ + languages: ['50 种语言'], + quality: 'high', + speed: 'slow', + recommended: false, + files: [ + { + role: 'detection', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_medium_det_onnx/resolve/c317b40325be40bfaaff58c8dcece2a075294f8a/inference.onnx', + size: 62_032_837, + sha256: + 'eb13b44b25bb36f89528b68720af8a61d9cf381176107f465db1757b65d086e1' + } + }, + { + role: 'recognition', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_medium_rec_onnx/resolve/db5d610d492a14e3c34dc1fd4e9339bd369f79e6/inference.onnx', + size: 76_554_979, + sha256: + '9c09abf0957f7968c7586464b7397b84ad2387a0497a351af40e9acc71b673ba' + } + }, + { + role: 'dictionary', + download: { + url: 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_medium_rec_onnx/resolve/db5d610d492a14e3c34dc1fd4e9339bd369f79e6/inference.yml', + size: 150_580, + sha256: + '991b700facf5b50a7de193468207d5f4255b538dde0d312ae3b7c7a9b6873129' + } + } + ] + }) + }) + + it('downloads, verifies, and loads OCR assets', async () => { + const { manager, modelBytes } = await createManager() + + await expect(manager.install('pp-ocrv6-tiny')).resolves.toMatchObject({ + id: 'pp-ocrv6-tiny', + source: 'download' + }) + await expect(manager.getStatus('pp-ocrv6-tiny')).resolves.toMatchObject({ + available: true, + verified: true + }) + const assets = await manager.getAssets('pp-ocrv6-tiny') + expect(new Uint8Array(assets.detection)).toEqual( + Uint8Array.from(modelBytes.detection) + ) + expect(new Uint8Array(assets.recognition)).toEqual( + Uint8Array.from(modelBytes.recognition) + ) + expect(new TextDecoder().decode(assets.dictionary)).toContain( + "!\n\"\n'\n" + ) + }) + + it('rejects an imported model whose hash does not match', async () => { + const { directory, manager, modelBytes } = await createManager() + const source = join(directory, 'manual-model') + await mkdir(source) + await Promise.all([ + writeFile(join(source, 'detection.onnx'), modelBytes.detection), + writeFile(join(source, 'recognition.onnx'), modelBytes.recognition), + writeFile(join(source, 'dictionary.yml'), 'tampered') + ]) + + await expect( + manager.registerLocalDirectory('pp-ocrv6-tiny', source) + ).rejects.toThrow('校验失败') + await expect(manager.getSnapshot()).resolves.toMatchObject({ + installed: [], + operations: [] + }) + }) + + it('round-trips a verified OCR model through an offline ZIP archive', async () => { + const { directory, manager, modelBytes } = await createManager() + const archive = join(directory, 'ocr-model.zip') + + await manager.install('pp-ocrv6-tiny') + await manager.exportArchive('pp-ocrv6-tiny', archive) + await manager.remove('pp-ocrv6-tiny') + + await expect( + manager.importArchive('pp-ocrv6-tiny', archive) + ).resolves.toMatchObject({ + id: 'pp-ocrv6-tiny', + source: 'local' + }) + const assets = await manager.getAssets('pp-ocrv6-tiny') + expect(new Uint8Array(assets.detection)).toEqual( + Uint8Array.from(modelBytes.detection) + ) + expect(new Uint8Array(assets.recognition)).toEqual( + Uint8Array.from(modelBytes.recognition) + ) + }) +}) + +describe('extractPaddleCharacterDictionary', () => { + it('converts Paddle YAML scalars into the line dictionary used by OCR', () => { + const dictionary = extractPaddleCharacterDictionary( + new TextDecoder().decode(dictionaryYaml()) + ) + expect(dictionary.startsWith("!\n\"\n'\n")).toBe(true) + expect(dictionary.split('\n')).toHaveLength(124) + }) +}) diff --git a/src/main/document-ocr-model-manager.ts b/src/main/document-ocr-model-manager.ts new file mode 100644 index 0000000..ae5bd6f --- /dev/null +++ b/src/main/document-ocr-model-manager.ts @@ -0,0 +1,899 @@ +import { createHash, randomUUID } from 'node:crypto' +import { + copyFile, + lstat, + mkdir, + open, + readFile, + readdir, + rename, + rm, + stat, + writeFile +} from 'node:fs/promises' +import { dirname, resolve } from 'node:path' +import { + documentOcrAssetsSchema, + documentOcrModelCatalogEntrySchema, + documentOcrModelSnapshotSchema, + documentParsingModelStatusSchema, + installedDocumentOcrModelSchema, + localOcrModelIdSchema, + type DocumentOcrAssets, + type DocumentOcrModelCatalogEntry, + type DocumentOcrModelFile, + type DocumentOcrModelOperation, + type DocumentOcrModelSnapshot, + type InstalledDocumentOcrModel +} from '../shared/document-parsing-contracts' +import { DOCUMENT_OCR_MODEL_CATALOG } from './document-ocr-model-catalog' +import { + exportModelArchive, + extractModelArchive +} from './model-archive' + +const DEFAULT_MAX_FILE_BYTES = 96 * 1024 * 1024 +const MANIFEST_FILE_NAME = 'manifest.json' +const MAX_REDIRECTS = 3 +const PARTIAL_SUFFIX = '.partial' +const MAXIMUM_ARCHIVE_BYTES = 512 * 1024 * 1024 +const ARCHIVE_OVERHEAD_BYTES = 1024 * 1024 +const executableExtensionPattern = + /\.(?:app|bat|bin|cmd|com|cpl|dll|dmg|exe|hta|inf|ins|iso|jar|js|jse|lnk|msi|msp|mst|pif|ps1|reg|scr|sh|sys|vb|vbe|vbs|ws|wsc|wsf|wsh)$/iu + +type ActiveOperation = { + controller: AbortController + progress: DocumentOcrModelOperation +} + +export type DocumentOcrModelManagerOptions = { + userDataDirectory: string + fetch: typeof fetch + catalog?: readonly DocumentOcrModelCatalogEntry[] + maxFileBytes?: number +} + +function abortError(): DOMException { + return new DOMException('The operation was aborted', 'AbortError') +} + +function ensureNotAborted(signal: AbortSignal): void { + if (signal.aborted) { + throw abortError() + } +} + +function cloneCatalogEntry( + entry: DocumentOcrModelCatalogEntry +): DocumentOcrModelCatalogEntry { + return documentOcrModelCatalogEntrySchema.parse(entry) +} + +function safeChild(parent: string, name: string): string { + const child = resolve(parent, name) + if (dirname(child) !== resolve(parent)) { + throw new Error('OCR 模型路径超出受管目录') + } + return child +} + +function validateDownloadUrl(value: string): URL { + const url = new URL(value) + if (url.protocol !== 'http:' && url.protocol !== 'https:') { + throw new Error('OCR 模型下载地址必须使用 HTTP 或 HTTPS') + } + return url +} + +function toArrayBuffer(buffer: Buffer): ArrayBuffer { + return Uint8Array.from(buffer).buffer +} + +async function hashFile( + path: string, + signal?: AbortSignal +): Promise<{ size: number; sha256: string }> { + const handle = await open(path, 'r') + const hash = createHash('sha256') + const buffer = Buffer.allocUnsafe(64 * 1024) + let size = 0 + try { + while (true) { + if (signal) { + ensureNotAborted(signal) + } + const { bytesRead } = await handle.read(buffer, 0, buffer.length) + if (bytesRead === 0) { + break + } + hash.update(buffer.subarray(0, bytesRead)) + size += bytesRead + } + } finally { + await handle.close() + } + return { size, sha256: hash.digest('hex') } +} + +function parseYamlScalar(value: string): string { + if (value.startsWith("'") && value.endsWith("'")) { + return value.slice(1, -1).replace(/''/gu, "'") + } + if (value.startsWith('"') && value.endsWith('"')) { + return JSON.parse(value) as string + } + return value +} + +export function extractPaddleCharacterDictionary(source: string): string { + const characters: string[] = [] + let readingDictionary = false + for (const line of source.replace(/\r/gu, '').split('\n')) { + if (line === ' character_dict:') { + readingDictionary = true + continue + } + if (!readingDictionary) { + continue + } + const match = /^ {2}- (.*)$/u.exec(line) + if (!match) { + break + } + const character = parseYamlScalar(match[1]!) + if (!character) { + throw new Error('OCR 字符字典包含空条目') + } + characters.push(character) + } + if (characters.length < 100) { + throw new Error('OCR 字符字典格式无效') + } + return `${characters.join('\n')}\n` +} + +export class DocumentOcrModelManager { + readonly rootDirectory: string + + private readonly transport: typeof fetch + private readonly catalog: DocumentOcrModelCatalogEntry[] + private readonly maxFileBytes: number + private readonly operations = new Map() + private readonly verifiedModels = new Map>() + + constructor(options: DocumentOcrModelManagerOptions) { + if (!options.userDataDirectory.trim()) { + throw new Error('userDataDirectory is required') + } + this.rootDirectory = resolve( + options.userDataDirectory, + 'models', + 'document-ocr' + ) + this.transport = options.fetch + this.catalog = (options.catalog ?? DOCUMENT_OCR_MODEL_CATALOG).map( + cloneCatalogEntry + ) + if ( + new Set(this.catalog.map((entry) => entry.id)).size !== + this.catalog.length + ) { + throw new Error('OCR 模型目录包含重复 ID') + } + this.maxFileBytes = options.maxFileBytes ?? DEFAULT_MAX_FILE_BYTES + if ( + !Number.isSafeInteger(this.maxFileBytes) || + this.maxFileBytes <= 0 || + this.maxFileBytes > 512 * 1024 * 1024 + ) { + throw new RangeError('maxFileBytes must be a positive safe integer') + } + } + + async getSnapshot(): Promise { + await this.ensureRoot() + return documentOcrModelSnapshotSchema.parse({ + rootDirectory: this.rootDirectory, + catalog: this.catalog.map(cloneCatalogEntry), + installed: await this.readInstalled(), + operations: [...this.operations.values()].map((operation) => ({ + ...operation.progress + })) + }) + } + + async getStatus( + modelId: string + ): Promise> { + const entry = this.requireCatalogEntry(modelId) + try { + await this.getVerifiedStatus(entry) + return documentParsingModelStatusSchema.parse({ + id: entry.id, + displayName: entry.displayName, + available: true, + verified: true, + runtime: entry.runtime, + detail: '模型已安装并通过 SHA-256 校验，可离线使用' + }) + } catch { + return documentParsingModelStatusSchema.parse({ + id: entry.id, + displayName: entry.displayName, + available: false, + verified: false, + runtime: entry.runtime, + detail: '模型尚未安装或校验失败，请从 ModelScope 下载' + }) + } + } + + getAssets(modelId: string): Promise { + return this.loadVerifiedAssets(this.requireCatalogEntry(modelId)) + } + + async install( + modelId: string, + externalSignal?: AbortSignal + ): Promise { + const entry = this.requireCatalogEntry(modelId) + const totalBytes = entry.files.reduce( + (total, file) => total + file.download.size, + 0 + ) + if (!Number.isSafeInteger(totalBytes)) { + throw new RangeError('OCR 模型总大小超出安全范围') + } + const operation = this.beginOperation(entry.id, 'download', totalBytes) + const detachAbort = this.attachExternalSignal( + externalSignal, + operation.controller + ) + let stagingDirectory: string | undefined + try { + await this.ensureRoot() + await this.assertNotInstalled(entry.id) + stagingDirectory = await this.createStagingDirectory(entry.id) + for (const file of entry.files) { + ensureNotAborted(operation.controller.signal) + operation.progress.phase = 'transferring' + operation.progress.currentFile = file.name + await this.downloadFile( + file, + safeChild(stagingDirectory, file.name), + operation, + operation.controller.signal + ) + } + operation.progress.phase = 'installing' + operation.progress.currentFile = null + const installed = await this.createInstalledManifest( + entry, + 'download', + stagingDirectory, + operation.controller.signal + ) + ensureNotAborted(operation.controller.signal) + await rename(stagingDirectory, this.modelDirectory(entry.id)) + stagingDirectory = undefined + this.verifiedModels.delete(entry.id) + return installed + } finally { + detachAbort() + this.operations.delete(entry.id) + if (stagingDirectory) { + await rm(stagingDirectory, { recursive: true, force: true }) + } + } + } + + async registerLocalDirectory( + modelId: string, + sourceDirectory: string, + externalSignal?: AbortSignal + ): Promise { + const entry = this.requireCatalogEntry(modelId) + const source = resolve(sourceDirectory) + const operation = this.beginOperation(entry.id, 'import', null) + const detachAbort = this.attachExternalSignal( + externalSignal, + operation.controller + ) + let stagingDirectory: string | undefined + try { + await this.ensureRoot() + await this.assertNotInstalled(entry.id) + await this.validateLocalDirectory( + source, + entry, + operation.controller.signal + ) + stagingDirectory = await this.createStagingDirectory(entry.id) + operation.progress.phase = 'transferring' + for (const file of entry.files) { + ensureNotAborted(operation.controller.signal) + operation.progress.currentFile = file.name + const sourceFile = safeChild(source, file.name) + const destination = safeChild(stagingDirectory, file.name) + await copyFile(sourceFile, destination) + operation.progress.completedBytes += + (await stat(destination)).size + } + operation.progress.totalBytes = + operation.progress.completedBytes + operation.progress.phase = 'installing' + operation.progress.currentFile = null + const installed = await this.createInstalledManifest( + entry, + 'local', + stagingDirectory, + operation.controller.signal + ) + ensureNotAborted(operation.controller.signal) + await rename(stagingDirectory, this.modelDirectory(entry.id)) + stagingDirectory = undefined + this.verifiedModels.delete(entry.id) + return installed + } finally { + detachAbort() + this.operations.delete(entry.id) + if (stagingDirectory) { + await rm(stagingDirectory, { recursive: true, force: true }) + } + } + } + + async exportArchive( + modelId: string, + destinationPath: string + ): Promise { + const entry = this.requireCatalogEntry(modelId) + await this.ensureRoot() + const installed = (await this.readInstalled()).find( + (model) => model.id === entry.id + ) + if (!installed) { + throw new Error('只能导出已安装的 OCR 模型') + } + const directory = this.modelDirectory(entry.id) + const files = [] + for (const expected of entry.files) { + const recorded = installed.files.find( + (file) => + file.name === expected.name && + file.role === expected.role + ) + if ( + !recorded || + recorded.size !== expected.download.size || + recorded.sha256 !== expected.download.sha256 + ) { + throw new Error(`OCR 模型文件校验失败：${expected.name}`) + } + files.push({ + name: expected.name, + role: expected.role, + size: recorded.size, + sha256: recorded.sha256 + }) + } + await exportModelArchive({ + destinationPath, + sourceDirectory: directory, + descriptor: { + kind: 'document-ocr', + modelId: entry.id, + displayName: entry.displayName, + files + } + }) + } + + async importArchive( + modelId: string, + archivePath: string + ): Promise { + const entry = this.requireCatalogEntry(modelId) + const expectedTotal = entry.files.reduce( + (total, file) => total + file.download.size, + 0 + ) + const operation = this.beginOperation( + entry.id, + 'import', + expectedTotal + ) + let stagingDirectory: string | undefined + try { + await this.ensureRoot() + await this.assertNotInstalled(entry.id) + stagingDirectory = await this.createStagingDirectory(entry.id) + operation.progress.phase = 'transferring' + const descriptor = await extractModelArchive({ + archivePath, + destinationDirectory: stagingDirectory, + expectedKind: 'document-ocr', + expectedModelId: entry.id, + expectedFiles: entry.files.map((file) => ({ + name: file.name, + role: file.role + })), + maximumArchiveBytes: Math.min( + MAXIMUM_ARCHIVE_BYTES, + expectedTotal + ARCHIVE_OVERHEAD_BYTES + ), + maximumFileBytes: this.maxFileBytes, + maximumTotalBytes: expectedTotal + ARCHIVE_OVERHEAD_BYTES, + signal: operation.controller.signal, + onProgress: (completedBytes) => { + operation.progress.completedBytes = completedBytes + } + }) + for (const expected of entry.files) { + const archived = descriptor.files.find( + (file) => + file.name === expected.name && + file.role === expected.role + ) + if ( + !archived || + archived.size !== expected.download.size || + archived.sha256 !== expected.download.sha256 + ) { + throw new Error( + `OCR 模型 ZIP 与当前模型目录不匹配：${expected.name}` + ) + } + } + operation.progress.phase = 'installing' + operation.progress.currentFile = null + const installed = installedDocumentOcrModelSchema.parse({ + id: entry.id, + displayName: entry.displayName, + source: 'local', + installedAt: new Date().toISOString(), + files: descriptor.files + }) + await writeFile( + safeChild(stagingDirectory, MANIFEST_FILE_NAME), + `${JSON.stringify(installed, null, 2)}\n`, + { encoding: 'utf8', flag: 'wx' } + ) + ensureNotAborted(operation.controller.signal) + await rename(stagingDirectory, this.modelDirectory(entry.id)) + stagingDirectory = undefined + this.verifiedModels.delete(entry.id) + return installed + } finally { + this.operations.delete(entry.id) + if (stagingDirectory) { + await rm(stagingDirectory, { recursive: true, force: true }) + } + } + } + + cancel(modelId: string): boolean { + const id = localOcrModelIdSchema.parse(modelId) + const operation = this.operations.get(id) + if (!operation) { + return false + } + operation.controller.abort() + return true + } + + async remove(modelId: string): Promise { + const id = localOcrModelIdSchema.parse(modelId) + this.cancel(id) + this.verifiedModels.delete(id) + await rm(this.modelDirectory(id), { + recursive: true, + force: true + }) + } + + dispose(): void { + for (const operation of this.operations.values()) { + operation.controller.abort() + } + this.operations.clear() + this.verifiedModels.clear() + } + + private async ensureRoot(): Promise { + await mkdir(this.rootDirectory, { recursive: true }) + } + + private modelDirectory(modelId: string): string { + return safeChild( + this.rootDirectory, + localOcrModelIdSchema.parse(modelId) + ) + } + + private requireCatalogEntry( + modelId: string + ): DocumentOcrModelCatalogEntry { + const id = localOcrModelIdSchema.parse(modelId) + const entry = this.catalog.find((candidate) => candidate.id === id) + if (!entry) { + throw new Error('未知的 OCR 模型') + } + return entry + } + + private beginOperation( + modelId: string, + kind: DocumentOcrModelOperation['kind'], + totalBytes: number | null + ): ActiveOperation { + if (this.operations.has(modelId)) { + throw new Error('该 OCR 模型已有进行中的操作') + } + const operation: ActiveOperation = { + controller: new AbortController(), + progress: { + modelId: localOcrModelIdSchema.parse(modelId), + kind, + phase: 'preparing', + currentFile: null, + completedBytes: 0, + totalBytes + } + } + this.operations.set(modelId, operation) + return operation + } + + private attachExternalSignal( + signal: AbortSignal | undefined, + controller: AbortController + ): () => void { + if (!signal) { + return () => undefined + } + const abort = (): void => controller.abort() + if (signal.aborted) { + controller.abort() + } else { + signal.addEventListener('abort', abort, { once: true }) + } + return () => signal.removeEventListener('abort', abort) + } + + private async assertNotInstalled(modelId: string): Promise { + try { + await lstat(this.modelDirectory(modelId)) + throw new Error('OCR 模型已安装') + } catch (error) { + if ( + error instanceof Error && + 'code' in error && + error.code === 'ENOENT' + ) { + return + } + throw error + } + } + + private async createStagingDirectory(modelId: string): Promise { + const directory = safeChild( + this.rootDirectory, + `.install-${modelId}-${randomUUID()}` + ) + await mkdir(directory, { recursive: false }) + return directory + } + + private async fetchFollowingRedirects( + initialUrl: string, + signal: AbortSignal + ): Promise { + let url = validateDownloadUrl(initialUrl) + for (let redirectCount = 0; ; redirectCount += 1) { + ensureNotAborted(signal) + const response = await this.transport(url, { + method: 'GET', + redirect: 'manual', + credentials: 'omit', + cache: 'no-store', + signal + }) + if ([301, 302, 303, 307, 308].includes(response.status)) { + if (redirectCount >= MAX_REDIRECTS) { + await response.body?.cancel().catch(() => undefined) + throw new Error('OCR 模型下载重定向次数过多') + } + const location = response.headers.get('location') + await response.body?.cancel().catch(() => undefined) + if (!location) { + throw new Error('OCR 模型下载重定向缺少地址') + } + url = validateDownloadUrl(new URL(location, url).toString()) + continue + } + return response + } + } + + private async downloadFile( + file: DocumentOcrModelFile, + destination: string, + operation: ActiveOperation, + signal: AbortSignal + ): Promise { + if (file.download.size > this.maxFileBytes) { + throw new RangeError(`OCR 模型文件过大：${file.name}`) + } + const response = await this.fetchFollowingRedirects( + file.download.url, + signal + ) + if (!response.ok) { + await response.body?.cancel().catch(() => undefined) + throw new Error(`OCR 模型下载失败：HTTP ${response.status}`) + } + if (!response.body) { + throw new Error('OCR 模型下载响应没有内容') + } + const declaredLength = response.headers.get('content-length') + if ( + declaredLength !== null && + Number(declaredLength) !== file.download.size + ) { + await response.body.cancel().catch(() => undefined) + throw new Error(`OCR 模型文件大小不匹配：${file.name}`) + } + + const partialPath = `${destination}${PARTIAL_SUFFIX}` + const handle = await open(partialPath, 'wx') + const reader = response.body.getReader() + const hash = createHash('sha256') + let written = 0 + try { + while (true) { + ensureNotAborted(signal) + const result = await reader.read() + if (result.done) { + break + } + written += result.value.byteLength + if ( + written > file.download.size || + written > this.maxFileBytes + ) { + await reader.cancel() + throw new RangeError(`OCR 模型文件过大：${file.name}`) + } + await handle.write(result.value) + hash.update(result.value) + operation.progress.completedBytes += result.value.byteLength + } + } catch (error) { + await reader.cancel().catch(() => undefined) + throw error + } finally { + await handle.close() + } + if ( + written !== file.download.size || + hash.digest('hex') !== file.download.sha256 + ) { + throw new Error(`OCR 模型文件校验失败：${file.name}`) + } + await rename(partialPath, destination) + } + + private async validateLocalDirectory( + sourceDirectory: string, + entry: DocumentOcrModelCatalogEntry, + signal: AbortSignal + ): Promise { + const sourceInfo = await lstat(sourceDirectory) + if (!sourceInfo.isDirectory() || sourceInfo.isSymbolicLink()) { + throw new Error('本地 OCR 模型来源必须是普通目录') + } + const entries = await readdir(sourceDirectory, { withFileTypes: true }) + for (const localEntry of entries) { + ensureNotAborted(signal) + if ( + localEntry.isSymbolicLink() || + executableExtensionPattern.test(localEntry.name) + ) { + throw new Error('本地 OCR 模型目录包含不安全文件') + } + } + for (const file of entry.files) { + ensureNotAborted(signal) + const path = safeChild(sourceDirectory, file.name) + const info = await lstat(path) + if (!info.isFile() || info.isSymbolicLink()) { + throw new Error(`OCR 模型文件必须是普通文件：${file.name}`) + } + const actual = await hashFile(path, signal) + if ( + actual.size !== file.download.size || + actual.sha256 !== file.download.sha256 + ) { + throw new Error(`本地 OCR 模型文件校验失败：${file.name}`) + } + } + } + + private async createInstalledManifest( + entry: DocumentOcrModelCatalogEntry, + source: InstalledDocumentOcrModel['source'], + stagingDirectory: string, + signal: AbortSignal + ): Promise { + const files = [] + for (const file of entry.files) { + ensureNotAborted(signal) + files.push({ + name: file.name, + role: file.role, + ...(await hashFile( + safeChild(stagingDirectory, file.name), + signal + )) + }) + } + const manifest = installedDocumentOcrModelSchema.parse({ + id: entry.id, + displayName: entry.displayName, + source, + installedAt: new Date().toISOString(), + files + }) + await writeFile( + safeChild(stagingDirectory, MANIFEST_FILE_NAME), + `${JSON.stringify(manifest, null, 2)}\n`, + { encoding: 'utf8', flag: 'wx' } + ) + return manifest + } + + private async readInstalled(): Promise { + const entries = await readdir(this.rootDirectory, { + withFileTypes: true + }) + const installed: InstalledDocumentOcrModel[] = [] + for (const entry of entries) { + if ( + !entry.isDirectory() || + entry.name.startsWith('.install-') || + !localOcrModelIdSchema.safeParse(entry.name).success + ) { + continue + } + try { + const manifest = installedDocumentOcrModelSchema.parse( + JSON.parse( + await readFile( + safeChild( + this.modelDirectory(entry.name), + MANIFEST_FILE_NAME + ), + 'utf8' + ) + ) as unknown + ) + if (manifest.id === entry.name) { + installed.push(manifest) + } + } catch { + // Ignore incomplete or externally modified model directories. + } + } + return installed + } + + private async readInstalledManifest( + entry: DocumentOcrModelCatalogEntry + ): Promise { + const directory = this.modelDirectory(entry.id) + const manifest = installedDocumentOcrModelSchema.parse( + JSON.parse( + await readFile( + safeChild(directory, MANIFEST_FILE_NAME), + 'utf8' + ) + ) as unknown + ) + if (manifest.id !== entry.id) { + throw new Error('OCR 模型清单 ID 不匹配') + } + return manifest + } + + private async verifyInstalledModel( + entry: DocumentOcrModelCatalogEntry + ): Promise { + const directory = this.modelDirectory(entry.id) + const manifest = await this.readInstalledManifest(entry) + for (const file of entry.files) { + const installed = manifest.files.find( + (candidate) => + candidate.name === file.name && + candidate.role === file.role + ) + const actual = await hashFile(safeChild(directory, file.name)) + if ( + !installed || + actual.size !== file.download.size || + actual.sha256 !== file.download.sha256 || + actual.size !== installed.size || + actual.sha256 !== installed.sha256 + ) { + throw new Error(`OCR 模型文件校验失败：${file.name}`) + } + } + } + + private getVerifiedStatus( + entry: DocumentOcrModelCatalogEntry + ): Promise { + let verification = this.verifiedModels.get(entry.id) + if (!verification) { + verification = this.verifyInstalledModel(entry).catch((error) => { + this.verifiedModels.delete(entry.id) + throw error + }) + this.verifiedModels.set(entry.id, verification) + } + return verification + } + + private async loadVerifiedAssets( + entry: DocumentOcrModelCatalogEntry + ): Promise { + const directory = this.modelDirectory(entry.id) + const manifest = await this.readInstalledManifest(entry) + const loaded = new Map< + DocumentOcrModelFile['role'], + ArrayBuffer + >() + for (const file of entry.files) { + const installed = manifest.files.find( + (candidate) => + candidate.name === file.name && + candidate.role === file.role + ) + const path = safeChild(directory, file.name) + const contents = await readFile(path) + const actual = { + size: contents.byteLength, + sha256: createHash('sha256').update(contents).digest('hex') + } + if ( + !installed || + actual.size !== file.download.size || + actual.sha256 !== file.download.sha256 || + actual.size !== installed.size || + actual.sha256 !== installed.sha256 + ) { + throw new Error(`OCR 模型文件校验失败：${file.name}`) + } + loaded.set( + file.role, + file.role === 'dictionary' + ? toArrayBuffer( + Buffer.from( + extractPaddleCharacterDictionary( + contents.toString('utf8') + ), + 'utf8' + ) + ) + : toArrayBuffer(contents) + ) + } + return documentOcrAssetsSchema.parse({ + modelId: entry.id, + detection: loaded.get('detection'), + recognition: loaded.get('recognition'), + dictionary: loaded.get('dictionary') + }) + } +} diff --git a/src/main/document-parsing-service.test.ts b/src/main/document-parsing-service.test.ts new file mode 100644 index 0000000..6fb139d --- /dev/null +++ b/src/main/document-parsing-service.test.ts @@ -0,0 +1,130 @@ +import { describe, expect, it, vi } from 'vitest' +import { + defaultDocumentParsingSettings +} from './document-parsing-settings-store' +import { DocumentParsingService } from './document-parsing-service' + +function createPdfFixture(text: string): Buffer { + const stream = `BT /F1 18 Tf 50 100 Td (${text}) Tj ET` + const objects = [ + '<< /Type /Catalog /Pages 2 0 R >>', + '<< /Type /Pages /Kids [3 0 R] /Count 1 >>', + '<< /Type /Page /Parent 2 0 R /MediaBox [0 0 300 200] /Resources << /Font << /F1 4 0 R >> >> /Contents 5 0 R >>', + '<< /Type /Font /Subtype /Type1 /BaseFont /Helvetica >>', + `<< /Length ${Buffer.byteLength(stream)} >>\nstream\n${stream}\nendstream` + ] + let content = '%PDF-1.4\n' + const offsets = [0] + for (const [index, object] of objects.entries()) { + offsets.push(Buffer.byteLength(content)) + content += `${index + 1} 0 obj\n${object}\nendobj\n` + } + const xrefOffset = Buffer.byteLength(content) + content += `xref\n0 ${objects.length + 1}\n` + content += '0000000000 65535 f \n' + content += offsets + .slice(1) + .map((offset) => `${String(offset).padStart(10, '0')} 00000 n \n`) + .join('') + content += `trailer\n<< /Size ${objects.length + 1} /Root 1 0 R >>\n` + content += `startxref\n${xrefOffset}\n%%EOF\n` + return Buffer.from(content) +} + +function createService(overrides?: { + settings?: Partial +}) { + const settings = { + ...defaultDocumentParsingSettings, + ...overrides?.settings + } + const recognize = vi.fn(async () => ({ + requestId: crypto.randomUUID(), + sections: [ + { + locator: '第 1 页', + content: '扫描件识别正文', + confidence: 0.93 + } + ], + pageCount: 1, + warnings: [] + })) + const service = new DocumentParsingService( + { + get: vi.fn(async () => settings), + update: vi.fn(async () => settings) + } as never, + { + getStatus: vi.fn(async () => ({ + id: 'pp-ocrv6-tiny', + displayName: 'PP-OCRv6 Tiny', + available: true, + verified: true, + runtime: 'onnxruntime-web-wasm', + detail: '可用' + })) + } as never, + { recognize } as never + ) + return { recognize, service } +} + +describe('DocumentParsingService', () => { + it('keeps useful PDF text local without invoking OCR', async () => { + const { recognize, service } = createService() + + const parsed = await service.parse( + 'native.pdf', + createPdfFixture('Native PDF body text'), + 'knowledge-index' + ) + + expect(parsed.content).toContain('Native PDF body text') + expect(parsed.sections[0]?.method).toBe('native') + expect(recognize).not.toHaveBeenCalled() + }) + + it('uses OCR for a PDF without useful text', async () => { + const { recognize, service } = createService() + + const parsed = await service.parse( + 'scan.pdf', + createPdfFixture(''), + 'chat-attachment' + ) + + expect(parsed.content).toBe('扫描件识别正文') + expect(parsed.sections).toEqual([ + { + locator: '第 1 页', + content: '扫描件识别正文', + method: 'ocr', + confidence: 0.93 + } + ]) + expect(recognize).toHaveBeenCalledWith( + expect.objectContaining({ + fileName: 'scan.pdf', + modelId: 'pp-ocrv6-tiny', + mimeType: 'application/pdf', + pageNumbers: [1] + }) + ) + }) + + it('does not use OCR in a fast-text workflow', async () => { + const { recognize, service } = createService({ + settings: { chatWorkflow: 'fast-text' } + }) + + await expect( + service.parse( + 'scan.pdf', + createPdfFixture(''), + 'chat-attachment' + ) + ).rejects.toThrow('未启用 OCR') + expect(recognize).not.toHaveBeenCalled() + }) +}) diff --git a/src/main/document-parsing-service.ts b/src/main/document-parsing-service.ts new file mode 100644 index 0000000..dc637c2 --- /dev/null +++ b/src/main/document-parsing-service.ts @@ -0,0 +1,270 @@ +import { extname } from 'node:path' +import { + documentParsingDiagnosticSchema, + documentParsingSnapshotSchema, + type DocumentParsingDiagnostic, + type DocumentParsingPurpose, + type DocumentParsingSettings, + type DocumentParsingSnapshot +} from '../shared/document-parsing-contracts' +import type { DocumentOcrBroker } from './document-ocr-broker' +import type { DocumentOcrModelManager } from './document-ocr-model-manager' +import type { DocumentParsingSettingsStore } from './document-parsing-settings-store' +import { + DocumentTextUnavailableError, + extractPdfTextPages, + parseDocument, + type ParsedDocument, + type ParsedSection, + type PdfTextPage +} from './knowledge/document-parser' + +const minimumUsefulPdfCharacters = 12 +const maximumReplacementCharacterRatio = 0.08 + +export type ParseDocumentForPurpose = ( + name: string, + buffer: Buffer, + purpose: DocumentParsingPurpose, + signal?: AbortSignal +) => Promise + +function ensureNotAborted(signal?: AbortSignal): void { + if (signal?.aborted) { + throw signal.reason instanceof Error + ? signal.reason + : new Error('文档解析已取消') + } +} + +function hasUsefulText(content: string): boolean { + const compact = content.replace(/\s+/gu, '') + if (compact.length < minimumUsefulPdfCharacters) { + return false + } + const replacementCount = [...compact].filter( + (character) => character === '\uFFFD' + ).length + return replacementCount / compact.length <= + maximumReplacementCharacterRatio +} + +function effectiveOcrMode( + settings: DocumentParsingSettings, + purpose: DocumentParsingPurpose +): DocumentParsingSettings['pdfOcrMode'] { + if ( + (purpose === 'chat-attachment' && + settings.chatWorkflow === 'fast-text') || + (purpose === 'knowledge-index' && + settings.knowledgeWorkflow === 'fast-index') + ) { + return 'disabled' + } + if ( + (purpose === 'chat-attachment' && + settings.chatWorkflow === 'high-fidelity') || + (purpose === 'knowledge-index' && + settings.knowledgeWorkflow === 'high-fidelity') + ) { + return 'always' + } + return settings.pdfOcrMode +} + +function buildPdfDocument( + name: string, + sections: ParsedSection[], + pageCount: number, + warnings: string[] = [] +): ParsedDocument { + const content = sections + .map((section) => section.content) + .join('\n\n') + .slice(0, 5_000_000) + if (!content) { + throw new DocumentTextUnavailableError() + } + return { + title: name.replace(/\.[^.]+$/u, ''), + sourceFormat: '.pdf', + content, + sections, + pageCount, + warnings + } +} + +function nativePdfSections(pages: PdfTextPage[]): ParsedSection[] { + return pages + .filter((page) => page.content.length > 0) + .map((page) => ({ + locator: `第 ${page.pageNumber} 页`, + content: page.content, + method: 'native' as const + })) +} + +export class DocumentParsingService { + constructor( + private readonly settingsStore: DocumentParsingSettingsStore, + private readonly modelManager: DocumentOcrModelManager, + private readonly ocrBroker: DocumentOcrBroker + ) {} + + async snapshot(): Promise { + const settings = await this.settingsStore.get() + const [localOcr, ocrModels] = await Promise.all([ + this.modelManager.getStatus(settings.localOcrModelId), + this.modelManager.getSnapshot() + ]) + return documentParsingSnapshotSchema.parse({ + settings, + status: { + nativeParsingAvailable: true, + conversionAvailable: false, + localOcr + }, + ocrModels + }) + } + + async update(input: unknown): Promise { + await this.settingsStore.update(input) + return this.snapshot() + } + + parse: ParseDocumentForPurpose = async ( + name, + buffer, + purpose, + signal + ) => { + ensureNotAborted(signal) + if (extname(name).toLowerCase() !== '.pdf') { + return parseDocument(name, buffer) + } + + const settings = await this.settingsStore.get() + const pages = await extractPdfTextPages(buffer) + ensureNotAborted(signal) + const mode = effectiveOcrMode(settings, purpose) + const pagesWithoutUsefulText = pages + .filter((page) => !hasUsefulText(page.content)) + .map((page) => page.pageNumber) + const ocrPageNumbers = + mode === 'always' + ? pages.map((page) => page.pageNumber) + : mode === 'auto' + ? pagesWithoutUsefulText + : [] + + if (mode === 'disabled' || !settings.localOcrEnabled) { + const native = nativePdfSections(pages) + if (native.length > 0) { + return buildPdfDocument( + name, + native, + pages.length, + pagesWithoutUsefulText.length > 0 + ? ['部分页面没有有效文本，当前工作流未启用 OCR'] + : [] + ) + } + throw new DocumentTextUnavailableError( + 'PDF 没有可用文本层，当前工作流未启用 OCR' + ) + } + if (ocrPageNumbers.length === 0) { + return buildPdfDocument( + name, + nativePdfSections(pages), + pages.length + ) + } + if (pages.length > settings.maximumPages) { + throw new Error( + `PDF 共 ${pages.length} 页，超过本地 OCR 的 ${settings.maximumPages} 页限制` + ) + } + const modelStatus = await this.modelManager.getStatus( + settings.localOcrModelId + ) + if (!modelStatus.available || !modelStatus.verified) { + throw new Error(modelStatus.detail) + } + + const ocrRequest = { + modelId: settings.localOcrModelId, + fileName: name, + mimeType: 'application/pdf' as const, + data: Uint8Array.from(buffer).buffer, + maximumPages: settings.maximumPages, + pageNumbers: ocrPageNumbers, + pageTimeoutSeconds: settings.pageTimeoutSeconds + } + const ocr = await (signal + ? this.ocrBroker.recognize(ocrRequest, signal) + : this.ocrBroker.recognize(ocrRequest)) + ensureNotAborted(signal) + const ocrByLocator = new Map( + ocr.sections.map((section) => [section.locator, section]) + ) + const merged = pages.flatMap((page): ParsedSection[] => { + const locator = `第 ${page.pageNumber} 页` + const recognized = ocrByLocator.get(locator) + if ( + recognized && + (mode === 'always' || !hasUsefulText(page.content)) + ) { + return [ + { + locator, + content: recognized.content, + method: 'ocr', + confidence: recognized.confidence + } + ] + } + return page.content + ? [{ locator, content: page.content, method: 'native' }] + : [] + }) + return buildPdfDocument(name, merged, pages.length, ocr.warnings) + } + + async diagnose( + name: string, + buffer: Buffer, + purpose: DocumentParsingPurpose = 'diagnostic' + ): Promise { + const startedAt = Date.now() + const parsed = await this.parse(name, buffer, purpose) + const ocrPageCount = parsed.sections.filter( + (section) => section.method === 'ocr' + ).length + const nativePageCount = parsed.sections.filter( + (section) => section.method !== 'ocr' + ).length + return documentParsingDiagnosticSchema.parse({ + fileName: name, + sourceFormat: + parsed.sourceFormat.replace(/^\./u, '').toUpperCase() || 'UNKNOWN', + pageCount: + parsed.sourceFormat === '.pdf' + ? (parsed.pageCount ?? parsed.sections.length) + : 0, + ocrPageCount, + characterCount: parsed.content.length, + method: + ocrPageCount > 0 && nativePageCount > 0 + ? 'mixed' + : ocrPageCount > 0 + ? 'ocr' + : 'native', + durationMs: Date.now() - startedAt, + preview: parsed.content.slice(0, 2_000), + warnings: parsed.warnings + }) + } +} diff --git a/src/main/document-parsing-settings-store.test.ts b/src/main/document-parsing-settings-store.test.ts new file mode 100644 index 0000000..ccdfab8 --- /dev/null +++ b/src/main/document-parsing-settings-store.test.ts @@ -0,0 +1,120 @@ +import { + mkdtemp, + readFile, + readdir, + rm, + writeFile +} from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { afterEach, describe, expect, it } from 'vitest' +import { + defaultDocumentParsingSettings, + DocumentParsingSettingsStore +} from './document-parsing-settings-store' + +const temporaryDirectories: string[] = [] + +async function createStore(): Promise<{ + directory: string + filePath: string + store: DocumentParsingSettingsStore +}> { + const directory = await mkdtemp( + join(tmpdir(), 'goodbuddy-document-parsing-settings-') + ) + temporaryDirectories.push(directory) + const filePath = join(directory, 'document-parsing-settings.json') + return { + directory, + filePath, + store: new DocumentParsingSettingsStore(filePath) + } +} + +afterEach(async () => { + await Promise.all( + temporaryDirectories.splice(0).map((directory) => + rm(directory, { recursive: true, force: true }) + ) + ) +}) + +describe('DocumentParsingSettingsStore', () => { + it('returns local-first defaults without creating a file', async () => { + const { directory, store } = await createStore() + + await expect(store.get()).resolves.toEqual( + defaultDocumentParsingSettings + ) + await expect(readdir(directory)).resolves.toEqual([]) + }) + + it('persists a complete versioned settings document', async () => { + const { filePath, store } = await createStore() + const settings = { + ...defaultDocumentParsingSettings, + chatWorkflow: 'fast-text' as const, + maximumPages: 42 + } + + await expect(store.update(settings)).resolves.toEqual(settings) + expect(JSON.parse(await readFile(filePath, 'utf8'))).toEqual({ + version: 2, + ...settings + }) + await expect( + new DocumentParsingSettingsStore(filePath).get() + ).resolves.toEqual(settings) + }) + + it('migrates legacy cloud permissions to the local OCR provider', async () => { + const { filePath, store } = await createStore() + const { + ocrProvider: _ocrProvider, + ...legacySettings + } = defaultDocumentParsingSettings + void _ocrProvider + await writeFile( + filePath, + JSON.stringify({ + version: 1, + ...legacySettings, + chatCloudPermission: 'always', + knowledgeCloudPermission: 'never' + }), + 'utf8' + ) + + await expect(store.get()).resolves.toEqual( + defaultDocumentParsingSettings + ) + }) + + it('rejects incomplete or out-of-range settings', async () => { + const { directory, store } = await createStore() + + await expect(store.update({})).rejects.toThrow() + await expect( + store.update({ + ...defaultDocumentParsingSettings, + maximumPages: 0 + }) + ).rejects.toThrow() + await expect(readdir(directory)).resolves.toEqual([]) + }) + + it('isolates corrupt settings and restores defaults', async () => { + const { directory, filePath, store } = await createStore() + await writeFile(filePath, '{not-json', 'utf8') + + await expect(store.get()).resolves.toEqual( + defaultDocumentParsingSettings + ) + const entries = await readdir(directory) + expect(entries).toHaveLength(1) + expect(entries[0]).toMatch( + /^document-parsing-settings\.json\.corrupt-\d+-[a-f0-9]{12}$/u + ) + }) +}) diff --git a/src/main/document-parsing-settings-store.ts b/src/main/document-parsing-settings-store.ts new file mode 100644 index 0000000..803e00b --- /dev/null +++ b/src/main/document-parsing-settings-store.ts @@ -0,0 +1,181 @@ +import { randomBytes } from 'node:crypto' +import { + mkdir, + readFile, + rename, + rm, + writeFile +} from 'node:fs/promises' +import { dirname } from 'node:path' +import { z } from 'zod' +import { + documentParsingSettingsSchema, + documentParsingSettingsUpdateSchema, + type DocumentParsingSettings +} from '../shared/document-parsing-contracts' + +const CURRENT_SETTINGS_VERSION = 2 + +const storedDocumentParsingSettingsSchema = + documentParsingSettingsSchema + .extend({ + version: z.literal(CURRENT_SETTINGS_VERSION) + }) + .strict() + +type StoredDocumentParsingSettings = z.infer< + typeof storedDocumentParsingSettingsSchema +> + +const legacyDocumentParsingSettingsSchema = + documentParsingSettingsSchema + .omit({ ocrProvider: true }) + .extend({ + version: z.literal(1), + chatCloudPermission: z.enum(['ask', 'always', 'never']), + knowledgeCloudPermission: z.enum(['ask', 'always', 'never']) + }) + .strict() + +export const defaultDocumentParsingSettings: DocumentParsingSettings = { + chatWorkflow: 'auto', + knowledgeWorkflow: 'complete-index', + pdfOcrMode: 'auto', + ocrProvider: 'local', + localOcrEnabled: true, + localOcrModelId: 'pp-ocrv6-tiny', + maximumPages: 100, + ocrConcurrency: 1, + pageTimeoutSeconds: 60 +} + +function isMissingFile(error: unknown): boolean { + return ( + error !== null && + typeof error === 'object' && + 'code' in error && + error.code === 'ENOENT' + ) +} + +export class DocumentParsingSettingsStore { + private settings?: StoredDocumentParsingSettings + private updateQueue: Promise = Promise.resolve() + + constructor(private readonly filePath: string) {} + + private async isolateCorruptFile(): Promise { + const isolatedPath = + `${this.filePath}.corrupt-${Date.now()}-` + + randomBytes(6).toString('hex') + try { + await rename(this.filePath, isolatedPath) + } catch (error) { + if (!isMissingFile(error)) { + throw new Error('文档解析设置损坏且无法隔离', { + cause: error + }) + } + } + } + + private async loadStored(): Promise { + if (this.settings) { + return this.settings + } + try { + const contents = await readFile(this.filePath, 'utf8') + let parsed: unknown + try { + parsed = JSON.parse(contents) as unknown + } catch { + await this.isolateCorruptFile() + this.settings = { + version: CURRENT_SETTINGS_VERSION, + ...defaultDocumentParsingSettings + } + return this.settings + } + const result = + storedDocumentParsingSettingsSchema.safeParse(parsed) + if (!result.success) { + const legacy = + legacyDocumentParsingSettingsSchema.safeParse(parsed) + if (legacy.success) { + const { + version: _version, + chatCloudPermission: _chatCloudPermission, + knowledgeCloudPermission: _knowledgeCloudPermission, + ...settings + } = legacy.data + void _version + void _chatCloudPermission + void _knowledgeCloudPermission + this.settings = { + version: CURRENT_SETTINGS_VERSION, + ocrProvider: 'local', + ...settings + } + return this.settings + } + await this.isolateCorruptFile() + this.settings = { + version: CURRENT_SETTINGS_VERSION, + ...defaultDocumentParsingSettings + } + return this.settings + } + this.settings = result.data + } catch (error) { + if (!isMissingFile(error)) { + throw new Error('无法读取文档解析设置', { cause: error }) + } + this.settings = { + version: CURRENT_SETTINGS_VERSION, + ...defaultDocumentParsingSettings + } + } + return this.settings + } + + async get(): Promise { + const { version: _version, ...settings } = await this.loadStored() + void _version + return documentParsingSettingsSchema.parse(settings) + } + + update(input: unknown): Promise { + const operation = this.updateQueue.then(async () => { + const updates = documentParsingSettingsUpdateSchema.parse(input) + const next: StoredDocumentParsingSettings = { + version: CURRENT_SETTINGS_VERSION, + ...updates + } + await mkdir(dirname(this.filePath), { recursive: true }) + const temporaryPath = + `${this.filePath}.${process.pid}.` + + `${randomBytes(6).toString('hex')}.tmp` + try { + await writeFile( + temporaryPath, + `${JSON.stringify(next, null, 2)}\n`, + { + encoding: 'utf8', + mode: 0o600, + flag: 'wx' + } + ) + await rename(temporaryPath, this.filePath) + } finally { + await rm(temporaryPath, { force: true }) + } + this.settings = next + return this.get() + }) + this.updateQueue = operation.then( + () => undefined, + () => undefined + ) + return operation + } +} diff --git a/src/main/index.ts b/src/main/index.ts index 1a05d0d..0a55836 100644 --- a/src/main/index.ts +++ b/src/main/index.ts @@ -67,6 +67,10 @@ import { KnowledgeEmbeddingIndexRepository } from './knowledge/knowledge-embeddi import { GlobalTlsPolicy } from './global-tls-policy' import type { AgentRuntimeSelection } from '../shared/runtime-selection-contracts' import { waitForCleanup } from './shutdown' +import { DocumentParsingSettingsStore } from './document-parsing-settings-store' +import { DocumentOcrModelManager } from './document-ocr-model-manager' +import { DocumentOcrBroker } from './document-ocr-broker' +import { DocumentParsingService } from './document-parsing-service' const shortcut = 'CommandOrControl+Shift+Space' const mainModuleDirectory = dirname(fileURLToPath(import.meta.url)) @@ -98,6 +102,8 @@ let knowledgeGateway: KnowledgeMcpGateway | undefined let assistantDatabase: AssistantDatabase | undefined let browserService: BrowserService | undefined let globalTlsPolicy: GlobalTlsPolicy | undefined +let documentOcrBroker: DocumentOcrBroker | undefined +let documentOcrModelManager: DocumentOcrModelManager | undefined function createEmbeddingProvider( settings: ResolvedRuntimeSettings @@ -340,6 +346,20 @@ if (hasSingleInstanceLock) { const applicationSettingsStore = new ApplicationSettingsStore( join(app.getPath('userData'), 'application-settings.json') ) + const documentParsingSettingsStore = + new DocumentParsingSettingsStore( + join(app.getPath('userData'), 'document-parsing-settings.json') + ) + documentOcrModelManager = new DocumentOcrModelManager({ + userDataDirectory: app.getPath('userData'), + fetch: globalThis.fetch + }) + documentOcrBroker = new DocumentOcrBroker(mainWindow) + const documentParsingService = new DocumentParsingService( + documentParsingSettingsStore, + documentOcrModelManager, + documentOcrBroker + ) const versionChecker = new VersionChecker({ fetch: globalThis.fetch, currentVersion: app.getVersion(), @@ -362,7 +382,8 @@ if (hasSingleInstanceLock) { knowledgeService = new KnowledgeService({ databasePath: join(app.getPath('userData'), 'knowledge.sqlite'), managedRoot: join(app.getPath('userData'), 'knowledge'), - extractStructured: createModelGraphExtractor(settingsStore) + extractStructured: createModelGraphExtractor(settingsStore), + parseDocument: documentParsingService.parse }) await knowledgeService.initialize() const embeddingIndexCoordinator = new EmbeddingIndexCoordinator( @@ -459,7 +480,9 @@ if (hasSingleInstanceLock) { selectedRuntimeManager = new SelectedRuntimeManager( createSelectedRuntime ) - const contextManager = new ContextManager() + const contextManager = new ContextManager({ + parseDocument: documentParsingService.parse + }) const approvalBroker = new ToolApprovalBroker() const shortcutRegistered = globalShortcut.register(shortcut, () => { @@ -510,7 +533,10 @@ if (hasSingleInstanceLock) { selectedRuntimeManager, speechTranscriptionService, knowledgeGateway, - launchWechatSidecar + launchWechatSidecar, + documentParsingService, + documentOcrModelManager, + documentOcrBroker ) loadMainWindow(mainWindow) @@ -550,7 +576,9 @@ app.on('before-quit', (event) => { Promise.resolve().then(() => knowledgeGateway?.dispose()), Promise.resolve().then(() => knowledgeService?.dispose()), Promise.resolve().then(() => browserService?.dispose()), - Promise.resolve().then(() => globalTlsPolicy?.dispose()) + Promise.resolve().then(() => globalTlsPolicy?.dispose()), + Promise.resolve().then(() => documentOcrModelManager?.dispose()), + Promise.resolve().then(() => documentOcrBroker?.dispose()) ]) globalShortcut.unregisterAll() tray?.destroy() diff --git a/src/main/ipc.test.ts b/src/main/ipc.test.ts index 09cb131..bdc3d8d 100644 --- a/src/main/ipc.test.ts +++ b/src/main/ipc.test.ts @@ -24,6 +24,10 @@ const electronMocks = vi.hoisted(() => { canceled: true, filePaths: [] as string[] })), + showSaveDialog: vi.fn(async () => ({ + canceled: true, + filePath: undefined as string | undefined + })), openPath: vi.fn(async () => ''), showItemInFolder: vi.fn(), openExternal: vi.fn(async () => undefined) @@ -248,7 +252,8 @@ vi.mock('electron', () => ({ }, BrowserWindow: class {}, dialog: { - showOpenDialog: electronMocks.showOpenDialog + showOpenDialog: electronMocks.showOpenDialog, + showSaveDialog: electronMocks.showSaveDialog }, ipcMain: { handle: electronMocks.handle, @@ -290,6 +295,180 @@ vi.mock('./channels/channel-env', () => ({ ) })) +describe('registerIpcHandlers model ZIP dialogs', () => { + afterEach(() => { + electronMocks.handlers.clear() + vi.clearAllMocks() + }) + + it('imports and exports speech and OCR ZIPs through trusted dialogs', async () => { + const webContents = { + mainFrame: { url: 'file:///goodbuddy/index.html' }, + getURL: vi.fn(() => 'file:///goodbuddy/index.html'), + send: vi.fn() + } + const window = { + webContents, + isDestroyed: vi.fn(() => false), + isMaximized: vi.fn(() => false), + on: vi.fn(), + removeListener: vi.fn() + } + const event = { + sender: webContents, + senderFrame: webContents.mainFrame + } + const speechSnapshot = { + catalog: [], + installed: [], + operations: [] + } + const speechModelManager = { + rootDirectory: 'C:\\models\\speech', + importArchive: vi.fn(async () => speechSnapshot), + exportArchive: vi.fn(async () => undefined), + getSnapshot: vi.fn(async () => speechSnapshot), + cancel: vi.fn() + } + const ocrSnapshot = { + settings: {}, + models: { + catalog: [], + installed: [], + operations: [] + } + } + const documentParsingService = { + snapshot: vi.fn(async () => ocrSnapshot) + } + const documentOcrModelManager = { + importArchive: vi.fn(async () => undefined), + exportArchive: vi.fn(async () => undefined) + } + const dispose = registerIpcHandlers( + window as never, + { capability: 'text' } as never, + 'CommandOrControl+Shift+Space', + {} as never, + {} as never, + { clear: vi.fn() } as never, + {} as never, + { claimDueSchedules: vi.fn(() => []) } as never, + { clear: vi.fn() } as never, + {} as never, + vi.fn(async () => undefined), + undefined, + undefined, + undefined, + undefined, + undefined, + undefined, + speechModelManager as never, + undefined, + undefined, + undefined, + undefined, + undefined, + documentParsingService as never, + documentOcrModelManager as never + ) + + await expect( + electronMocks.handlers.get( + ipcChannels.speechModelsImportArchive + )?.(event, { modelId: 'speech-model' }) + ).resolves.toBeUndefined() + expect(speechModelManager.importArchive).not.toHaveBeenCalled() + + electronMocks.showOpenDialog.mockResolvedValueOnce({ + canceled: false, + filePaths: ['C:\\transfer\\speech.zip'] + }) + await expect( + electronMocks.handlers.get( + ipcChannels.speechModelsImportArchive + )?.(event, { modelId: 'speech-model' }) + ).resolves.toBe(speechSnapshot) + expect(electronMocks.showOpenDialog).toHaveBeenLastCalledWith( + window, + expect.objectContaining({ + properties: ['openFile'], + filters: [ + { + name: 'GoodBuddy 模型 ZIP', + extensions: ['zip'] + } + ] + }) + ) + expect(speechModelManager.importArchive).toHaveBeenCalledWith( + 'speech-model', + 'C:\\transfer\\speech.zip' + ) + + electronMocks.showSaveDialog.mockResolvedValueOnce({ + canceled: false, + filePath: 'C:\\transfer\\speech-model' + }) + await expect( + electronMocks.handlers.get( + ipcChannels.speechModelsExportArchive + )?.(event, { modelId: 'speech-model' }) + ).resolves.toBe(speechSnapshot) + expect(speechModelManager.exportArchive).toHaveBeenCalledWith( + 'speech-model', + 'C:\\transfer\\speech-model.zip' + ) + + electronMocks.showOpenDialog.mockResolvedValueOnce({ + canceled: false, + filePaths: ['C:\\transfer\\ocr.zip'] + }) + await expect( + electronMocks.handlers.get( + ipcChannels.documentOcrModelsImportArchive + )?.(event, { modelId: 'ocr-model' }) + ).resolves.toBe(ocrSnapshot) + expect(documentOcrModelManager.importArchive).toHaveBeenCalledWith( + 'ocr-model', + 'C:\\transfer\\ocr.zip' + ) + + electronMocks.showSaveDialog.mockResolvedValueOnce({ + canceled: false, + filePath: 'C:\\transfer\\ocr-model.ZIP' + }) + await expect( + electronMocks.handlers.get( + ipcChannels.documentOcrModelsExportArchive + )?.(event, { modelId: 'ocr-model' }) + ).resolves.toBe(ocrSnapshot) + expect(documentOcrModelManager.exportArchive).toHaveBeenCalledWith( + 'ocr-model', + 'C:\\transfer\\ocr-model.ZIP' + ) + + await expect( + electronMocks.handlers.get( + ipcChannels.speechModelsExportArchive + )?.( + { + sender: {}, + senderFrame: webContents.mainFrame + }, + { modelId: 'speech-model' } + ) + ).rejects.toThrow('拒绝来自未知窗口的 IPC 请求') + await expect( + electronMocks.handlers.get( + ipcChannels.documentOcrModelsImportArchive + )?.(event, {}) + ).rejects.toThrow() + + await dispose() + }) +}) + describe('registerIpcHandlers connection tests', () => { afterEach(() => { electronMocks.handlers.clear() diff --git a/src/main/ipc.ts b/src/main/ipc.ts index d22bcd4..b366636 100644 --- a/src/main/ipc.ts +++ b/src/main/ipc.ts @@ -73,6 +73,12 @@ import { embeddingIndexJobRequestSchema, embeddingSettingsSnapshotSchema } from '../shared/embedding-contracts' +import { + documentOcrModelActionInputSchema, + documentOcrFailureSchema, + documentOcrResultSchema, + documentParsingSettingsUpdateSchema +} from '../shared/document-parsing-contracts' import { agentRuntimeSelectionSchema, type AgentRuntimeSelection @@ -178,6 +184,9 @@ import type { VersionChecker } from './version-checker' import type { SpeechModelManager } from './speech/speech-model-manager' import type { SpeechTranscriptionService } from './speech/speech-transcription-service' import type { EmbeddingIndexCoordinator } from './knowledge/embedding-index-coordinator' +import type { DocumentParsingService } from './document-parsing-service' +import type { DocumentOcrModelManager } from './document-ocr-model-manager' +import type { DocumentOcrBroker } from './document-ocr-broker' import { OpenAIEmbeddingClient } from './knowledge/openai-embedding-client' import { magicNotePlainText, @@ -321,6 +330,17 @@ const taskStatusRequestSchema = z status: z.enum(['completed', 'cancelled']) }) .strict() + +const modelArchiveDialogFilters = [ + { + name: 'GoodBuddy 模型 ZIP', + extensions: ['zip'] + } +] + +function ensureZipExtension(path: string): string { + return extname(path).toLowerCase() === '.zip' ? path : `${path}.zip` +} const expertUpdateRequestSchema = z .object({ expertId: assistantIdSchema, @@ -575,7 +595,10 @@ export function registerIpcHandlers( selectedRuntimes?: SelectedRuntimeResolver, speechTranscriptionService?: SpeechTranscriptionService, knowledgeGateway?: KnowledgeMcpGateway, - launchWechatSidecar?: WechatSidecarLauncher + launchWechatSidecar?: WechatSidecarLauncher, + documentParsingService?: DocumentParsingService, + documentOcrModelManager?: DocumentOcrModelManager, + documentOcrBroker?: DocumentOcrBroker ): () => Promise { const activeRequests = new Map() const pendingAgentQuestions = new Map< @@ -2463,6 +2486,233 @@ export function registerIpcHandlers( } ) + ipcMain.handle(ipcChannels.documentParsingGet, (event) => { + assertTrustedSender(event, window) + if (!documentParsingService) { + throw new Error('文档解析设置服务不可用') + } + return documentParsingService.snapshot() + }) + + ipcMain.handle( + ipcChannels.documentParsingUpdate, + (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentParsingService) { + throw new Error('文档解析设置服务不可用') + } + return documentParsingService.update( + documentParsingSettingsUpdateSchema.parse(input) + ) + } + ) + + ipcMain.handle( + ipcChannels.documentParsingTest, + async (event) => { + assertTrustedSender(event, window) + if (!documentParsingService) { + throw new Error('文档解析设置服务不可用') + } + const result = await dialog.showOpenDialog(window, { + title: '选择测试文档', + properties: ['openFile'], + filters: [ + { + name: '支持的文档', + extensions: supportedDocumentExtensions.map((extension) => + extension.slice(1) + ) + } + ] + }) + const selectedPath = result.filePaths[0] + if (result.canceled || !selectedPath) { + return undefined + } + try { + const canonicalPath = await realpath(selectedPath) + const fileStat = await stat(canonicalPath) + if (!fileStat.isFile() || fileStat.size > 20 * 1024 * 1024) { + throw new Error('测试文档必须小于 20MB 且不能是目录') + } + return documentParsingService.diagnose( + basename(canonicalPath), + await readFile(canonicalPath) + ) + } catch (error) { + if (error instanceof Error && !('code' in error)) { + throw error + } + throw new Error('无法读取测试文档，请检查文件权限和状态', { + cause: error + }) + } + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsInstall, + (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager || !documentParsingService) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + return trackExecution( + documentOcrModelManager + .install(modelId) + .then(() => documentParsingService.snapshot()) + ) + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsCancel, + (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + return documentOcrModelManager.cancel(modelId) + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsRemove, + async (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager || !documentParsingService) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + await documentOcrModelManager.remove(modelId) + return documentParsingService.snapshot() + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsImportArchive, + async (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager || !documentParsingService) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + const result = await dialog.showOpenDialog(window, { + title: '导入 OCR 模型 ZIP', + properties: ['openFile'], + filters: modelArchiveDialogFilters + }) + const archivePath = result.filePaths[0] + if (result.canceled || !archivePath) { + return undefined + } + return trackExecution( + documentOcrModelManager + .importArchive(modelId, archivePath) + .then(() => documentParsingService.snapshot()) + ) + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsExportArchive, + async (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager || !documentParsingService) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + const result = await dialog.showSaveDialog(window, { + title: '导出 OCR 模型 ZIP', + defaultPath: `${modelId}.zip`, + filters: modelArchiveDialogFilters + }) + if (result.canceled || !result.filePath) { + return undefined + } + const destination = ensureZipExtension(result.filePath) + await documentOcrModelManager.exportArchive( + modelId, + destination + ) + return documentParsingService.snapshot() + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsOpenRepository, + async (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + const snapshot = await documentOcrModelManager.getSnapshot() + const entry = snapshot.catalog.find( + (candidate) => candidate.id === modelId + ) + if (!entry) { + throw new Error('未知的 OCR 模型') + } + await shell.openExternal(entry.repositoryUrl) + } + ) + + ipcMain.handle( + ipcChannels.documentOcrModelsOpenDirectory, + async (event) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager) { + throw new Error('本地 OCR 模型服务不可用') + } + await documentOcrModelManager.getSnapshot() + const error = await shell.openPath( + documentOcrModelManager.rootDirectory + ) + if (error) { + throw new Error('无法打开 OCR 模型目录') + } + } + ) + + ipcMain.handle( + ipcChannels.documentParsingOcrAssets, + (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrModelManager) { + throw new Error('本地 OCR 模型服务不可用') + } + const { modelId } = + documentOcrModelActionInputSchema.parse(input) + return documentOcrModelManager.getAssets(modelId) + } + ) + + ipcMain.handle( + ipcChannels.documentParsingOcrRespond, + (event, input: unknown) => { + assertTrustedSender(event, window) + if (!documentOcrBroker) { + throw new Error('本地 OCR 任务服务不可用') + } + const result = documentOcrResultSchema.safeParse(input) + documentOcrBroker.respond( + result.success + ? result.data + : documentOcrFailureSchema.parse(input) + ) + } + ) + ipcMain.handle(ipcChannels.versionCheck, async (event) => { assertTrustedSender(event, window) if (!versionChecker) { @@ -2610,7 +2860,7 @@ export function registerIpcHandlers( ) ipcMain.handle( - ipcChannels.speechModelsImportLocal, + ipcChannels.speechModelsImportArchive, async (event, input: unknown) => { assertTrustedSender(event, window) if (!speechModelManager) { @@ -2618,20 +2868,44 @@ export function registerIpcHandlers( } const { modelId } = speechModelActionInputSchema.parse(input) const result = await dialog.showOpenDialog(window, { - properties: ['openDirectory'] + title: '导入语音模型 ZIP', + properties: ['openFile'], + filters: modelArchiveDialogFilters }) - const directory = result.filePaths[0] - if (result.canceled || !directory) { + const archivePath = result.filePaths[0] + if (result.canceled || !archivePath) { return undefined } return trackExecution( speechModelManager - .registerLocalDirectory(modelId, directory) + .importArchive(modelId, archivePath) .then(() => speechModelManager.getSnapshot()) ) } ) + ipcMain.handle( + ipcChannels.speechModelsExportArchive, + async (event, input: unknown) => { + assertTrustedSender(event, window) + if (!speechModelManager) { + throw new Error('语音模型服务不可用') + } + const { modelId } = speechModelActionInputSchema.parse(input) + const result = await dialog.showSaveDialog(window, { + title: '导出语音模型 ZIP', + defaultPath: `${modelId}.zip`, + filters: modelArchiveDialogFilters + }) + if (result.canceled || !result.filePath) { + return undefined + } + const destination = ensureZipExtension(result.filePath) + await speechModelManager.exportArchive(modelId, destination) + return speechModelManager.getSnapshot() + } + ) + ipcMain.handle( ipcChannels.speechModelsOpenRepository, async (event, input: unknown) => { diff --git a/src/main/knowledge/document-parser-electron-runtime.test.ts b/src/main/knowledge/document-parser-electron-runtime.test.ts new file mode 100644 index 0000000..5ccc99d --- /dev/null +++ b/src/main/knowledge/document-parser-electron-runtime.test.ts @@ -0,0 +1,50 @@ +import { beforeEach, describe, expect, it, vi } from 'vitest' + +const getDocument = vi.hoisted(() => vi.fn()) + +vi.mock('pdfjs-dist/legacy/build/pdf.mjs', () => ({ + getDocument +})) + +import { extractPdfTextPages } from './document-parser' + +describe('PDF extraction in Electron main', () => { + beforeEach(() => { + getDocument.mockReset() + }) + + it('disables PDF.js DOM factories for headless text extraction', async () => { + const cleanup = vi.fn() + const destroy = vi.fn(async () => undefined) + getDocument.mockReturnValue({ + promise: Promise.resolve({ + numPages: 1, + getPage: vi.fn(async () => ({ + getTextContent: vi.fn(async () => ({ + items: [{ str: 'PDF body text' }] + })), + cleanup + })) + }), + destroy + }) + + await expect( + extractPdfTextPages(Buffer.from('synthetic PDF')) + ).resolves.toEqual([ + { + pageNumber: 1, + content: 'PDF body text' + } + ]) + expect(getDocument).toHaveBeenCalledWith({ + data: expect.any(Uint8Array), + disableFontFace: true, + isOffscreenCanvasSupported: false, + useSystemFonts: false, + useWorkerFetch: false + }) + expect(cleanup).toHaveBeenCalledOnce() + expect(destroy).toHaveBeenCalledOnce() + }) +}) diff --git a/src/main/knowledge/document-parser.ts b/src/main/knowledge/document-parser.ts index 5bd223e..31e3135 100644 --- a/src/main/knowledge/document-parser.ts +++ b/src/main/knowledge/document-parser.ts @@ -5,12 +5,17 @@ import { extname } from 'node:path' export type ParsedSection = { locator: string content: string + method?: 'native' | 'ocr' | 'converted' | 'vision' + confidence?: number } export type ParsedDocument = { title: string + sourceFormat: string content: string sections: ParsedSection[] + warnings: string[] + pageCount?: number } export type DocumentChunk = { @@ -160,12 +165,43 @@ function parseOfficeArchive( } async function parsePdf(buffer: Buffer): Promise { + const pages = await extractPdfTextPages(buffer) + return pages + .filter((page) => page.content.length > 0) + .map((page) => ({ + locator: `第 ${page.pageNumber} 页`, + content: page.content + })) +} + +export type PdfTextPage = { + pageNumber: number + content: string +} + +export class DocumentTextUnavailableError extends Error { + constructor(message = '文档中没有可索引的文本内容') { + super(message) + this.name = 'DocumentTextUnavailableError' + } +} + +export async function extractPdfTextPages( + buffer: Buffer +): Promise { const pdfjs = await import('pdfjs-dist/legacy/build/pdf.mjs') const loadingTask = pdfjs.getDocument({ - data: new Uint8Array(buffer) + data: new Uint8Array(buffer), + // Electron's main process identifies itself as process.type === + // "browser", so PDF.js otherwise selects DOM font factories even + // though no document exists there. + disableFontFace: true, + isOffscreenCanvasSupported: false, + useSystemFonts: false, + useWorkerFetch: false }) const document = await loadingTask.promise - const sections: ParsedSection[] = [] + const pages: PdfTextPage[] = [] try { for (let pageNumber = 1; pageNumber <= document.numPages; pageNumber += 1) { const page = await document.getPage(pageNumber) @@ -175,18 +211,13 @@ async function parsePdf(buffer: Buffer): Promise { .join(' ') .replace(/\s+/g, ' ') .trim() - if (content) { - sections.push({ - locator: `第 ${pageNumber} 页`, - content - }) - } + pages.push({ pageNumber, content }) page.cleanup() } } finally { await loadingTask.destroy() } - return sections + return pages } export async function parseDocument( @@ -227,12 +258,14 @@ export async function parseDocument( .join('\n\n') .slice(0, maximumExtractedCharacters) if (!content) { - throw new Error('文档中没有可索引的文本内容') + throw new DocumentTextUnavailableError() } return { title: name.replace(/\.[^.]+$/, ''), + sourceFormat: extension || 'unknown', content, - sections + sections, + warnings: [] } } diff --git a/src/main/knowledge/knowledge-service.ts b/src/main/knowledge/knowledge-service.ts index 6f7d929..060d8eb 100644 --- a/src/main/knowledge/knowledge-service.ts +++ b/src/main/knowledge/knowledge-service.ts @@ -18,7 +18,12 @@ import { relative, resolve } from 'node:path' -import { chunkDocument, parseDocument, supportedDocumentExtensions } from './document-parser' +import { + chunkDocument, + parseDocument, + supportedDocumentExtensions, + type ParsedDocument +} from './document-parser' import { classifyEmbeddingError } from './embedding-errors' import { extractKnowledgeGraph, @@ -99,6 +104,12 @@ export type KnowledgeServiceOptions = { urlImporter?: UrlImporter embeddingProvider?: EmbeddingProvider embeddingBatchSize?: number + parseDocument?: ( + name: string, + buffer: Buffer, + purpose: 'knowledge-index', + signal?: AbortSignal + ) => Promise } const supportedExtensions = new Set(supportedDocumentExtensions) @@ -118,6 +129,9 @@ export class KnowledgeService { private readonly managedRoot: string private readonly extractStructured?: ExtractStructured private readonly urlImporter: UrlImporter + private readonly documentParser: NonNullable< + KnowledgeServiceOptions['parseDocument'] + > private embeddingProvider?: EmbeddingProvider private readonly embeddingBatchSize: number private readonly watchers = new Map() @@ -131,6 +145,9 @@ export class KnowledgeService { this.managedRoot = resolve(options.managedRoot) this.extractStructured = options.extractStructured this.urlImporter = options.urlImporter ?? new UrlImporter() + this.documentParser = + options.parseDocument ?? + ((name, buffer) => parseDocument(name, buffer)) this.embeddingProvider = options.embeddingProvider const embeddingBatchSize = options.embeddingBatchSize ?? 16 if ( @@ -839,9 +856,11 @@ export class KnowledgeService { }) continue } - const parsed = await parseDocument( + const parsed = await this.documentParser( basename(file.absolutePath), - buffer + buffer, + 'knowledge-index', + this.lifecycleController.signal ) this.updateKnowledgeTask(parsingTask.id, { progress: 75, diff --git a/src/main/model-archive.test.ts b/src/main/model-archive.test.ts new file mode 100644 index 0000000..d9c5efb --- /dev/null +++ b/src/main/model-archive.test.ts @@ -0,0 +1,211 @@ +import { createHash } from 'node:crypto' +import { + mkdtemp, + mkdir, + readFile, + rm, + writeFile +} from 'node:fs/promises' +import { tmpdir } from 'node:os' +import { join } from 'node:path' +import { zipSync } from 'fflate' +import { afterEach, describe, expect, it } from 'vitest' +import { + exportModelArchive, + extractModelArchive +} from './model-archive' + +const temporaryDirectories: string[] = [] + +async function temporaryDirectory(): Promise { + const directory = await mkdtemp( + join(tmpdir(), 'goodbuddy-model-archive-') + ) + temporaryDirectories.push(directory) + return directory +} + +function sha256(value: Uint8Array): string { + return createHash('sha256').update(value).digest('hex') +} + +afterEach(async () => { + await Promise.all( + temporaryDirectories.splice(0).map((directory) => + rm(directory, { recursive: true, force: true }) + ) + ) +}) + +describe('model archive', () => { + it('exports and extracts only declared verified model files', async () => { + const directory = await temporaryDirectory() + const source = join(directory, 'source') + const extracted = join(directory, 'extracted') + const archive = join(directory, 'model.zip') + await Promise.all([mkdir(source), mkdir(extracted)]) + const model = Buffer.from('verified model bytes') + const tokens = Buffer.from('verified tokens') + await Promise.all([ + writeFile(join(source, 'model.onnx'), model), + writeFile(join(source, 'tokens.txt'), tokens), + writeFile(join(source, 'ignored.txt'), 'not exported'), + writeFile(archive, 'archive selected for replacement') + ]) + + await exportModelArchive({ + destinationPath: archive, + sourceDirectory: source, + descriptor: { + kind: 'speech', + modelId: 'test-model', + displayName: 'Test model', + files: [ + { + name: 'model.onnx', + role: 'model', + size: model.byteLength, + sha256: sha256(model) + }, + { + name: 'tokens.txt', + role: 'tokens', + size: tokens.byteLength, + sha256: sha256(tokens) + } + ] + } + }) + + await expect( + extractModelArchive({ + archivePath: archive, + destinationDirectory: extracted, + expectedKind: 'speech', + expectedModelId: 'test-model', + expectedFiles: [ + { name: 'model.onnx', role: 'model' }, + { name: 'tokens.txt', role: 'tokens' } + ], + maximumArchiveBytes: 1024 * 1024, + maximumFileBytes: 1024, + maximumTotalBytes: 2048 + }) + ).resolves.toMatchObject({ + kind: 'speech', + modelId: 'test-model' + }) + await expect(readFile(join(extracted, 'model.onnx'))).resolves.toEqual( + model + ) + await expect(readFile(join(extracted, 'tokens.txt'))).resolves.toEqual( + tokens + ) + }) + + it('preserves an existing archive when source verification fails', async () => { + const directory = await temporaryDirectory() + const source = join(directory, 'source') + const archive = join(directory, 'model.zip') + await mkdir(source) + const model = Buffer.from('changed model') + await Promise.all([ + writeFile(join(source, 'model.onnx'), model), + writeFile(archive, 'existing archive') + ]) + + await expect( + exportModelArchive({ + destinationPath: archive, + sourceDirectory: source, + descriptor: { + kind: 'speech', + modelId: 'test-model', + displayName: 'Test model', + files: [ + { + name: 'model.onnx', + role: 'model', + size: model.byteLength, + sha256: 'a'.repeat(64) + } + ] + } + }) + ).rejects.toThrow('模型文件校验失败') + await expect(readFile(archive, 'utf8')).resolves.toBe( + 'existing archive' + ) + }) + + it('rejects path traversal and undeclared archive entries', async () => { + const directory = await temporaryDirectory() + const archive = join(directory, 'unsafe.zip') + const extracted = join(directory, 'extracted') + await mkdir(extracted) + await writeFile( + archive, + zipSync({ + '../model.onnx': Buffer.from('unsafe') + }) + ) + + await expect( + extractModelArchive({ + archivePath: archive, + destinationDirectory: extracted, + expectedKind: 'speech', + expectedModelId: 'test-model', + expectedFiles: [{ name: 'model.onnx', role: 'model' }], + maximumArchiveBytes: 1024 * 1024, + maximumFileBytes: 1024, + maximumTotalBytes: 1024 + }) + ).rejects.toThrow() + }) + + it('rejects an archive whose manifest model ID does not match', async () => { + const directory = await temporaryDirectory() + const archive = join(directory, 'mismatch.zip') + const extracted = join(directory, 'extracted') + await mkdir(extracted) + const model = Buffer.from('model') + await writeFile( + archive, + zipSync({ + 'goodbuddy-model.json': Buffer.from( + JSON.stringify({ + format: 'goodbuddy-model-archive', + version: 1, + kind: 'speech', + modelId: 'other-model', + displayName: 'Other model', + exportedAt: '2026-08-11T00:00:00.000Z', + files: [ + { + name: 'model.onnx', + role: 'model', + size: model.byteLength, + sha256: sha256(model) + } + ] + }) + ), + 'model.onnx': model + }) + ) + + await expect( + extractModelArchive({ + archivePath: archive, + destinationDirectory: extracted, + expectedKind: 'speech', + expectedModelId: 'test-model', + expectedFiles: [{ name: 'model.onnx', role: 'model' }], + maximumArchiveBytes: 1024 * 1024, + maximumFileBytes: 1024, + maximumTotalBytes: 1024 + }) + ).rejects.toThrow('模型 ID 不匹配') + }) +}) diff --git a/src/main/model-archive.ts b/src/main/model-archive.ts new file mode 100644 index 0000000..7f30172 --- /dev/null +++ b/src/main/model-archive.ts @@ -0,0 +1,617 @@ +import { createHash, randomUUID } from 'node:crypto' +import { + lstat, + open, + readFile, + rename, + rm, + type FileHandle +} from 'node:fs/promises' +import { dirname, resolve } from 'node:path' +import { + Unzip, + UnzipInflate, + UnzipPassThrough, + Zip, + ZipPassThrough +} from 'fflate' +import { z } from 'zod' + +const ARCHIVE_MANIFEST_NAME = 'goodbuddy-model.json' +const ARCHIVE_FORMAT = 'goodbuddy-model-archive' +const ARCHIVE_VERSION = 1 +const MAXIMUM_ARCHIVE_ENTRIES = 40 +const MAXIMUM_MANIFEST_BYTES = 256 * 1024 + +const archiveFileNameSchema = z + .string() + .min(1) + .max(255) + .regex(/^[^/\\:\0]+$/u) + +const modelArchiveFileSchema = z + .object({ + name: archiveFileNameSchema, + role: z.string().trim().min(1).max(64), + size: z.number().int().positive().safe(), + sha256: z.string().regex(/^[a-f0-9]{64}$/u) + }) + .strict() + +const modelArchiveDescriptorSchema = z + .object({ + kind: z.enum(['speech', 'document-ocr']), + modelId: z + .string() + .min(1) + .max(96) + .regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/u), + displayName: z.string().trim().min(1).max(120), + files: z.array(modelArchiveFileSchema).min(1).max(32) + }) + .strict() + +const modelArchiveManifestSchema = modelArchiveDescriptorSchema + .extend({ + format: z.literal(ARCHIVE_FORMAT), + version: z.literal(ARCHIVE_VERSION), + exportedAt: z.string().datetime() + }) + .strict() + .superRefine((manifest, context) => { + if ( + new Set(manifest.files.map((file) => file.name.toLowerCase())) + .size !== manifest.files.length + ) { + context.addIssue({ + code: 'custom', + path: ['files'], + message: '模型 ZIP 清单包含重复文件' + }) + } + }) + +export type ModelArchiveKind = z.infer< + typeof modelArchiveManifestSchema +>['kind'] + +export type ModelArchiveFile = z.infer + +export type ModelArchiveDescriptor = { + kind: ModelArchiveKind + modelId: string + displayName: string + files: ModelArchiveFile[] +} + +export type ModelArchiveExpectedFile = { + name: string + role: string +} + +type ExportModelArchiveOptions = { + destinationPath: string + sourceDirectory: string + descriptor: ModelArchiveDescriptor +} + +type ExtractModelArchiveOptions = { + archivePath: string + destinationDirectory: string + expectedKind: ModelArchiveKind + expectedModelId: string + expectedFiles: ModelArchiveExpectedFile[] + maximumArchiveBytes: number + maximumFileBytes: number + maximumTotalBytes: number + signal?: AbortSignal + onProgress?: (completedBytes: number) => void +} + +function safeChild(parent: string, name: string): string { + const child = resolve(parent, name) + if (dirname(child) !== resolve(parent)) { + throw new Error('模型 ZIP 路径超出临时目录') + } + return child +} + +function ensureArchiveName(name: string): string { + return archiveFileNameSchema.parse(name) +} + +function ensureUniqueFiles(files: ModelArchiveExpectedFile[]): void { + const names = files.map((file) => ensureArchiveName(file.name)) + if (new Set(names.map((name) => name.toLowerCase())).size !== names.length) { + throw new Error('模型目录包含重复文件名') + } +} + +async function hashFile(path: string): Promise { + const handle = await open(path, 'r') + const hash = createHash('sha256') + const buffer = Buffer.allocUnsafe(64 * 1024) + try { + while (true) { + const { bytesRead } = await handle.read(buffer, 0, buffer.length) + if (bytesRead === 0) { + break + } + hash.update(buffer.subarray(0, bytesRead)) + } + } finally { + await handle.close() + } + return hash.digest('hex') +} + +function checkedLimit(value: number, label: string): number { + if (!Number.isSafeInteger(value) || value <= 0) { + throw new RangeError(`${label}无效`) + } + return value +} + +function ensureNotAborted(signal?: AbortSignal): void { + if (signal?.aborted) { + throw signal.reason instanceof Error + ? signal.reason + : new Error('模型 ZIP 导入已取消') + } +} + +async function pushFileIntoArchive( + archive: Zip, + file: ModelArchiveFile, + sourcePath: string, + waitForOutput: () => Promise +): Promise { + const input = new ZipPassThrough(ensureArchiveName(file.name)) + archive.add(input) + const sourceInfo = await lstat(sourcePath) + if (!sourceInfo.isFile() || sourceInfo.isSymbolicLink()) { + throw new Error(`模型文件不可导出：${file.name}`) + } + const handle = await open(sourcePath, 'r') + const buffer = Buffer.allocUnsafe(64 * 1024) + const hash = createHash('sha256') + let size = 0 + try { + const openedInfo = await handle.stat() + if ( + !openedInfo.isFile() || + openedInfo.dev !== sourceInfo.dev || + openedInfo.ino !== sourceInfo.ino + ) { + throw new Error(`模型文件在打开前已发生变化：${file.name}`) + } + while (true) { + const { bytesRead } = await handle.read(buffer, 0, buffer.length) + if (bytesRead === 0) { + break + } + const chunk = buffer.subarray(0, bytesRead) + hash.update(chunk) + size += bytesRead + input.push(Uint8Array.from(chunk)) + await waitForOutput() + } + if (size !== file.size || hash.digest('hex') !== file.sha256) { + throw new Error(`模型文件校验失败：${file.name}`) + } + input.push(new Uint8Array(), true) + await waitForOutput() + } finally { + await handle.close() + } +} + +async function pushBytesIntoArchive( + archive: Zip, + name: string, + value: Uint8Array, + waitForOutput: () => Promise +): Promise { + const input = new ZipPassThrough(ensureArchiveName(name)) + archive.add(input) + input.push(value, true) + await waitForOutput() +} + +async function replaceArchiveFile( + partialPath: string, + destinationPath: string +): Promise { + const backupPath = `${destinationPath}.${randomUUID()}.backup` + let movedExistingFile = false + try { + try { + await rename(destinationPath, backupPath) + movedExistingFile = true + const existingInfo = await lstat(backupPath) + if (!existingInfo.isFile() || existingInfo.isSymbolicLink()) { + throw new Error('模型 ZIP 导出目标必须是普通文件') + } + } catch (error) { + if ((error as NodeJS.ErrnoException).code !== 'ENOENT') { + throw error + } + } + await rename(partialPath, destinationPath) + if (movedExistingFile) { + await rm(backupPath, { force: true }).catch(() => undefined) + } + } catch (error) { + if (movedExistingFile) { + await rm(destinationPath, { force: true }).catch(() => undefined) + await rename(backupPath, destinationPath).catch(() => undefined) + } + throw error + } +} + +export async function exportModelArchive( + options: ExportModelArchiveOptions +): Promise { + const descriptor = modelArchiveDescriptorSchema.parse( + options.descriptor + ) + ensureUniqueFiles(descriptor.files) + const sourceDirectory = resolve(options.sourceDirectory) + const destinationPath = resolve(options.destinationPath) + const partialPath = `${destinationPath}.${randomUUID()}.partial` + const output = await open(partialPath, 'wx') + let writeChain = Promise.resolve() + let archiveError: Error | undefined + let resolveFinished: (() => void) | undefined + let rejectFinished: ((error: Error) => void) | undefined + const finished = new Promise((resolvePromise, rejectPromise) => { + resolveFinished = resolvePromise + rejectFinished = rejectPromise + }) + const archive = new Zip((error, data, final) => { + if (error) { + archiveError = error + rejectFinished?.(error) + return + } + writeChain = writeChain.then(async () => { + if (data.byteLength > 0) { + await output.write(data) + } + }) + if (final) { + void writeChain.then(resolveFinished, rejectFinished) + } + }) + const waitForOutput = async (): Promise => { + await writeChain + if (archiveError) { + throw archiveError + } + } + try { + const manifest = modelArchiveManifestSchema.parse({ + format: ARCHIVE_FORMAT, + version: ARCHIVE_VERSION, + kind: descriptor.kind, + modelId: descriptor.modelId, + displayName: descriptor.displayName, + exportedAt: new Date().toISOString(), + files: descriptor.files + }) + await pushBytesIntoArchive( + archive, + ARCHIVE_MANIFEST_NAME, + Buffer.from(`${JSON.stringify(manifest, null, 2)}\n`, 'utf8'), + waitForOutput + ) + for (const file of descriptor.files) { + await pushFileIntoArchive( + archive, + file, + safeChild(sourceDirectory, file.name), + waitForOutput + ) + } + archive.end() + await finished + await output.sync() + await output.close() + await replaceArchiveFile(partialPath, destinationPath) + } catch (error) { + archive.terminate() + await output.close().catch(() => undefined) + await rm(partialPath, { force: true }) + throw error + } +} + +function closeHandle(handle: FileHandle): Promise { + return handle.close().catch(() => undefined) +} + +export async function extractModelArchive( + options: ExtractModelArchiveOptions +): Promise { + ensureNotAborted(options.signal) + const maximumArchiveBytes = checkedLimit( + options.maximumArchiveBytes, + '模型 ZIP 大小限制' + ) + const maximumFileBytes = checkedLimit( + options.maximumFileBytes, + '模型文件大小限制' + ) + const maximumTotalBytes = checkedLimit( + options.maximumTotalBytes, + '模型展开大小限制' + ) + const expectedFiles = options.expectedFiles.map((file) => ({ + name: ensureArchiveName(file.name), + role: file.role + })) + ensureUniqueFiles(expectedFiles) + const allowedNames = new Set([ + ARCHIVE_MANIFEST_NAME, + ...expectedFiles.map((file) => file.name) + ]) + const source = resolve(options.archivePath) + let sourceInfo + try { + sourceInfo = await lstat(source) + } catch (error) { + throw new Error('无法读取模型 ZIP', { cause: error }) + } + if ( + !sourceInfo.isFile() || + sourceInfo.isSymbolicLink() || + sourceInfo.size <= 0 || + sourceInfo.size > maximumArchiveBytes + ) { + throw new Error('模型 ZIP 必须是大小合规的普通文件') + } + + let input: FileHandle | undefined + try { + input = await open(source, 'r') + const openedInfo = await input.stat() + if ( + !openedInfo.isFile() || + openedInfo.size !== sourceInfo.size || + openedInfo.dev !== sourceInfo.dev || + openedInfo.ino !== sourceInfo.ino + ) { + await input.close() + throw new Error('模型 ZIP 在打开前已发生变化') + } + } catch (error) { + await input?.close().catch(() => undefined) + if (error instanceof Error && error.message.startsWith('模型 ZIP')) { + throw error + } + throw new Error('无法读取模型 ZIP', { cause: error }) + } + if (!input) { + throw new Error('无法读取模型 ZIP') + } + + const destination = resolve(options.destinationDirectory) + const seenNames = new Set() + const openHandles = new Set() + const completions: Promise[] = [] + const pendingWrites = new Set>() + let entryCount = 0 + let totalBytes = 0 + let completedModelBytes = 0 + let fatalError: Error | undefined + const fail = (error: unknown): Error => { + const resolvedError = + error instanceof Error ? error : new Error('模型 ZIP 已损坏') + fatalError ??= resolvedError + return resolvedError + } + const unzip = new Unzip((file) => { + try { + entryCount += 1 + if ( + entryCount > MAXIMUM_ARCHIVE_ENTRIES || + entryCount > allowedNames.size + ) { + throw new Error('模型 ZIP 包含过多条目') + } + const name = ensureArchiveName(file.name) + const key = name.toLowerCase() + if (seenNames.has(key)) { + throw new Error('模型 ZIP 包含重复条目') + } + seenNames.add(key) + if (!allowedNames.has(name)) { + throw new Error(`模型 ZIP 包含未声明文件：${name}`) + } + const entryMaximum = + name === ARCHIVE_MANIFEST_NAME + ? MAXIMUM_MANIFEST_BYTES + : maximumFileBytes + if ( + file.originalSize !== undefined && + (file.originalSize <= 0 || + file.originalSize > entryMaximum || + totalBytes + file.originalSize > maximumTotalBytes) + ) { + throw new Error(`模型 ZIP 条目大小超出限制：${name}`) + } + const handlePromise = open( + safeChild(destination, name), + 'wx' + ).then((handle) => { + openHandles.add(handle) + return handle + }) + let written = 0 + let writeChain = Promise.resolve() + let resolveEntry: (() => void) | undefined + let rejectEntry: ((error: Error) => void) | undefined + const completion = new Promise((resolveEntryPromise, rejectEntryPromise) => { + resolveEntry = resolveEntryPromise + rejectEntry = rejectEntryPromise + }) + completions.push(completion) + file.ondata = (error, data, final) => { + if (error) { + rejectEntry?.(fail(error)) + return + } + if (fatalError) { + file.terminate() + rejectEntry?.(fatalError) + return + } + if (options.signal?.aborted) { + file.terminate() + rejectEntry?.( + fail( + options.signal.reason instanceof Error + ? options.signal.reason + : new Error('模型 ZIP 导入已取消') + ) + ) + return + } + written += data.byteLength + totalBytes += data.byteLength + if (name !== ARCHIVE_MANIFEST_NAME) { + completedModelBytes += data.byteLength + options.onProgress?.(completedModelBytes) + } + if ( + written > entryMaximum || + totalBytes > maximumTotalBytes + ) { + file.terminate() + rejectEntry?.( + fail(new Error(`模型 ZIP 条目大小超出限制：${name}`)) + ) + return + } + writeChain = writeChain.then(async () => { + const handle = await handlePromise + if (data.byteLength > 0) { + await handle.write(data) + } + }) + const pendingWrite = writeChain + pendingWrites.add(pendingWrite) + void pendingWrite.then( + () => pendingWrites.delete(pendingWrite), + () => pendingWrites.delete(pendingWrite) + ) + if (final) { + void writeChain.then(async () => { + const handle = await handlePromise + openHandles.delete(handle) + await closeHandle(handle) + resolveEntry?.() + }, (writeError: unknown) => { + rejectEntry?.(fail(writeError)) + }) + } + } + file.start() + } catch (error) { + file.terminate() + fail(error) + } + }) + unzip.register(UnzipPassThrough) + unzip.register(UnzipInflate) + + const buffer = Buffer.allocUnsafe(16 * 1024) + try { + while (true) { + ensureNotAborted(options.signal) + if (fatalError) { + throw fatalError + } + const { bytesRead } = await input.read(buffer, 0, buffer.length) + if (bytesRead === 0) { + unzip.push(new Uint8Array(), true) + break + } + unzip.push( + Uint8Array.from(buffer.subarray(0, bytesRead)), + false + ) + await Promise.all([...pendingWrites]) + } + await Promise.all(completions) + if (fatalError) { + throw fatalError + } + } catch (error) { + throw fail(error) + } finally { + await input.close() + await Promise.all( + [...openHandles].map((handle) => closeHandle(handle)) + ) + } + + if ( + seenNames.size !== allowedNames.size || + [...allowedNames].some( + (name) => !seenNames.has(name.toLowerCase()) + ) + ) { + throw new Error('模型 ZIP 缺少必需文件') + } + + let manifest + try { + manifest = modelArchiveManifestSchema.parse( + JSON.parse( + await readFile( + safeChild(destination, ARCHIVE_MANIFEST_NAME), + 'utf8' + ) + ) as unknown + ) + } catch { + throw new Error('模型 ZIP 清单无效') + } + if ( + manifest.kind !== options.expectedKind || + manifest.modelId !== options.expectedModelId + ) { + throw new Error('模型 ZIP 类型或模型 ID 不匹配') + } + if ( + manifest.files.length !== expectedFiles.length || + expectedFiles.some((expected) => { + const archived = manifest.files.find( + (file) => file.name === expected.name + ) + return !archived || archived.role !== expected.role + }) + ) { + throw new Error('模型 ZIP 清单与当前模型目录不匹配') + } + for (const archived of manifest.files) { + const path = safeChild(destination, archived.name) + const metadata = await lstat(path) + if ( + !metadata.isFile() || + metadata.isSymbolicLink() || + metadata.size !== archived.size || + (await hashFile(path)) !== archived.sha256 + ) { + throw new Error(`模型 ZIP 文件校验失败：${archived.name}`) + } + } + return { + kind: manifest.kind, + modelId: manifest.modelId, + displayName: manifest.displayName, + files: manifest.files + } +} diff --git a/src/main/speech/speech-model-manager.test.ts b/src/main/speech/speech-model-manager.test.ts index cb6d9a0..f9d24af 100644 --- a/src/main/speech/speech-model-manager.test.ts +++ b/src/main/speech/speech-model-manager.test.ts @@ -373,6 +373,47 @@ describe('SpeechModelManager downloads', () => { operations: [] }) }) + + it('round-trips a verified model through an offline ZIP archive', async () => { + const userData = await temporaryDirectory() + const modelBytes = new TextEncoder().encode('verified model bytes') + const tokenBytes = new TextEncoder().encode('verified tokens') + const catalog = downloadableCatalog(modelBytes, tokenBytes) + const manager = new SpeechModelManager({ + userDataDirectory: userData, + catalog, + fetch: vi.fn(async (input) => { + const bytes = String(input).endsWith('model.onnx') + ? modelBytes + : tokenBytes + return new Response(bytes, { + headers: { 'content-length': String(bytes.byteLength) } + }) + }) + }) + const archive = join(userData, 'speech-model.zip') + + await manager.install('download-test-model') + await manager.exportArchive('download-test-model', archive) + await manager.remove('download-test-model') + + await expect( + manager.importArchive('download-test-model', archive) + ).resolves.toMatchObject({ + id: 'download-test-model', + source: 'local', + files: [ + { + name: 'model.onnx', + sha256: sha256(modelBytes) + }, + { + name: 'tokens.txt', + sha256: sha256(tokenBytes) + } + ] + }) + }) }) describe('SpeechModelManager local import', () => { diff --git a/src/main/speech/speech-model-manager.ts b/src/main/speech/speech-model-manager.ts index f8d18f6..545777a 100644 --- a/src/main/speech/speech-model-manager.ts +++ b/src/main/speech/speech-model-manager.ts @@ -25,12 +25,18 @@ import { type SpeechModelSnapshot } from '../../shared/speech-model-contracts' import { SPEECH_MODEL_CATALOG } from './speech-model-catalog' +import { + exportModelArchive, + extractModelArchive +} from '../model-archive' const DEFAULT_MAX_FILE_BYTES = 2 * 1024 * 1024 * 1024 const MAX_REDIRECTS = 3 const MANIFEST_FILE_NAME = 'manifest.json' const SELECTION_FILE_NAME = '.selection.json' const PARTIAL_SUFFIX = '.partial' +const MAXIMUM_ARCHIVE_BYTES = 4 * 1024 * 1024 * 1024 - 1 +const ARCHIVE_OVERHEAD_BYTES = 1024 * 1024 const selectionSchema = z .object({ @@ -380,6 +386,143 @@ export class SpeechModelManager { } } + async exportArchive( + modelId: string, + destinationPath: string + ): Promise { + const entry = this.requireCatalogEntry(modelId) + await this.ensureRoot() + const installed = (await this.readInstalled()).find( + (model) => model.id === entry.id + ) + if (!installed) { + throw new Error('只能导出已安装的语音模型') + } + const directory = this.modelDirectory(entry.id) + const files = [] + for (const expected of entry.files) { + const recorded = installed.files.find( + (file) => + file.name === expected.name && file.role === expected.role + ) + if ( + !recorded || + recorded.size <= 0 || + recorded.size > this.maxFileBytes || + (expected.download && + (recorded.size !== expected.download.size || + recorded.sha256 !== expected.download.sha256)) + ) { + throw new Error(`语音模型文件不可导出：${expected.name}`) + } + files.push({ + name: expected.name, + role: expected.role, + size: recorded.size, + sha256: recorded.sha256 + }) + } + await exportModelArchive({ + destinationPath, + sourceDirectory: directory, + descriptor: { + kind: 'speech', + modelId: entry.id, + displayName: entry.displayName, + files + } + }) + } + + async importArchive( + modelId: string, + archivePath: string + ): Promise { + const entry = this.requireCatalogEntry(modelId) + const expectedTotal = entry.files.reduce( + (total, file) => + total + (file.download?.size ?? this.maxFileBytes), + 0 + ) + const maximumTotalBytes = Math.min( + MAXIMUM_ARCHIVE_BYTES, + expectedTotal + ARCHIVE_OVERHEAD_BYTES + ) + const operation = this.beginOperation( + entry.id, + 'import', + expectedTotal + ) + let stagingDirectory: string | undefined + try { + await this.ensureRoot() + await this.assertNotInstalled(entry.id) + stagingDirectory = await this.createStagingDirectory(entry.id) + operation.progress.phase = 'transferring' + const descriptor = await extractModelArchive({ + archivePath, + destinationDirectory: stagingDirectory, + expectedKind: 'speech', + expectedModelId: entry.id, + expectedFiles: entry.files.map((file) => ({ + name: file.name, + role: file.role + })), + maximumArchiveBytes: Math.min( + MAXIMUM_ARCHIVE_BYTES, + maximumTotalBytes + ARCHIVE_OVERHEAD_BYTES + ), + maximumFileBytes: this.maxFileBytes, + maximumTotalBytes, + signal: operation.controller.signal, + onProgress: (completedBytes) => { + operation.progress.completedBytes = completedBytes + } + }) + for (const expected of entry.files) { + const archived = descriptor.files.find( + (file) => + file.name === expected.name && + file.role === expected.role + ) + if ( + !archived || + archived.size > this.maxFileBytes || + (expected.download && + (archived.size !== expected.download.size || + archived.sha256 !== expected.download.sha256)) + ) { + throw new Error( + `语音模型 ZIP 与当前模型目录不匹配：${expected.name}` + ) + } + } + operation.progress.phase = 'installing' + operation.progress.currentFile = null + const installed = installedSpeechModelSchema.parse({ + id: entry.id, + displayName: entry.displayName, + source: 'local', + installedAt: new Date().toISOString(), + files: descriptor.files + }) + await writeFile( + safeChild(stagingDirectory, MANIFEST_FILE_NAME), + `${JSON.stringify(installed, null, 2)}\n`, + { encoding: 'utf8', flag: 'wx' } + ) + ensureNotAborted(operation.controller.signal) + await rename(stagingDirectory, this.modelDirectory(entry.id)) + stagingDirectory = undefined + return installed + } finally { + this.operations.delete(entry.id) + if (stagingDirectory) { + await rm(stagingDirectory, { recursive: true, force: true }) + } + } + } + private async ensureRoot(): Promise { await mkdir(this.rootDirectory, { recursive: true }) } diff --git a/src/preload/index.ts b/src/preload/index.ts index 8fa9776..0de2755 100644 --- a/src/preload/index.ts +++ b/src/preload/index.ts @@ -75,6 +75,15 @@ import type { EmbeddingIndexStatus, EmbeddingSettingsSnapshot } from '../shared/embedding-contracts' +import type { + DocumentOcrAssets, + DocumentOcrFailure, + DocumentOcrRequest, + DocumentOcrResult, + DocumentParsingDiagnostic, + DocumentParsingSettings, + DocumentParsingSnapshot +} from '../shared/document-parsing-contracts' import type { AgentRuntimeSelection } from '../shared/runtime-selection-contracts' import type { WeixinBindingSnapshot } from '../shared/weixin-channel-contracts' import type { RemoteChannelActivity } from '../shared/remote-channel-contracts' @@ -345,9 +354,14 @@ const desktopApi: DesktopApi = { ipcChannels.speechModelsSelect, { modelId } ) as Promise, - importLocalDirectory: (modelId: string) => + importArchive: (modelId: string) => ipcRenderer.invoke( - ipcChannels.speechModelsImportLocal, + ipcChannels.speechModelsImportArchive, + { modelId } + ) as Promise, + exportArchive: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.speechModelsExportArchive, { modelId } ) as Promise, openRepository: async (modelId: string) => { @@ -403,6 +417,94 @@ const desktopApi: DesktopApi = { ) } }, + documentParsing: { + getSnapshot: () => + ipcRenderer.invoke( + ipcChannels.documentParsingGet + ) as Promise, + update: (input: DocumentParsingSettings) => + ipcRenderer.invoke( + ipcChannels.documentParsingUpdate, + input + ) as Promise, + test: () => + ipcRenderer.invoke( + ipcChannels.documentParsingTest + ) as Promise, + installOcrModel: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentOcrModelsInstall, + { modelId } + ) as Promise, + cancelOcrModelOperation: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentOcrModelsCancel, + { modelId } + ) as Promise, + removeOcrModel: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentOcrModelsRemove, + { modelId } + ) as Promise, + importOcrModelArchive: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentOcrModelsImportArchive, + { modelId } + ) as Promise, + exportOcrModelArchive: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentOcrModelsExportArchive, + { modelId } + ) as Promise, + openOcrModelRepository: async (modelId: string) => { + await ipcRenderer.invoke( + ipcChannels.documentOcrModelsOpenRepository, + { modelId } + ) + }, + openOcrModelsDirectory: async () => { + await ipcRenderer.invoke( + ipcChannels.documentOcrModelsOpenDirectory + ) + }, + getOcrAssets: (modelId: string) => + ipcRenderer.invoke( + ipcChannels.documentParsingOcrAssets, + { modelId } + ) as Promise, + respondOcr: async ( + response: DocumentOcrResult | DocumentOcrFailure + ) => { + await ipcRenderer.invoke( + ipcChannels.documentParsingOcrRespond, + response + ) + }, + onOcrRequest: (listener) => { + const handler = ( + _event: Electron.IpcRendererEvent, + request: DocumentOcrRequest + ): void => listener(request) + ipcRenderer.on(ipcChannels.documentParsingOcrRequest, handler) + return () => + ipcRenderer.removeListener( + ipcChannels.documentParsingOcrRequest, + handler + ) + }, + onOcrCancel: (listener) => { + const handler = ( + _event: Electron.IpcRendererEvent, + requestId: string + ): void => listener(requestId) + ipcRenderer.on(ipcChannels.documentParsingOcrCancel, handler) + return () => + ipcRenderer.removeListener( + ipcChannels.documentParsingOcrCancel, + handler + ) + } + }, projects: { list: (includeArchived = false) => ipcRenderer.invoke( diff --git a/src/preload/preload-sandbox.test.ts b/src/preload/preload-sandbox.test.ts index 905f7d3..46af679 100644 --- a/src/preload/preload-sandbox.test.ts +++ b/src/preload/preload-sandbox.test.ts @@ -36,4 +36,17 @@ describe('sandboxed preload', () => { /(?:setComputerCapability|BrowserProfile).{0,80}(?:executablePath|command|env|args)/su ) }) + + it('exposes model ZIP dialogs without renderer-controlled paths', () => { + const source = readFileSync( + join(process.cwd(), 'src', 'preload', 'index.ts'), + 'utf8' + ) + expect(source).toContain('importArchive: (modelId: string)') + expect(source).toContain('exportArchive: (modelId: string)') + expect(source).toContain('importOcrModelArchive: (modelId: string)') + expect(source).toContain('exportOcrModelArchive: (modelId: string)') + expect(source).not.toContain('importLocalDirectory:') + expect(source).not.toContain('importOcrModel:') + }) }) diff --git a/src/renderer/src/DocumentParsingSettingsSection.test.tsx b/src/renderer/src/DocumentParsingSettingsSection.test.tsx new file mode 100644 index 0000000..8ebdad4 --- /dev/null +++ b/src/renderer/src/DocumentParsingSettingsSection.test.tsx @@ -0,0 +1,377 @@ +import { + cleanup, + fireEvent, + render, + screen, + waitFor +} from '@testing-library/react' +import { afterEach, beforeEach, describe, expect, it, vi } from 'vitest' +import type { + DocumentParsingSettings, + DocumentParsingSnapshot +} from '../../shared/document-parsing-contracts' +import { changeUiLocale } from './i18n' +import { DocumentParsingSettingsSection } from './DocumentParsingSettingsSection' + +const settings: DocumentParsingSettings = { + chatWorkflow: 'auto', + knowledgeWorkflow: 'complete-index', + pdfOcrMode: 'auto', + ocrProvider: 'local', + localOcrEnabled: true, + localOcrModelId: 'pp-ocrv6-tiny', + maximumPages: 100, + ocrConcurrency: 1, + pageTimeoutSeconds: 60 +} + +const modelEntry = { + id: 'pp-ocrv6-tiny' as const, + displayName: 'PP-OCRv6 Tiny', + description: '轻量中文 OCR 模型', + languages: ['中文', '英语'], + runtime: 'onnxruntime-web-wasm' as const, + quality: 'basic' as const, + speed: 'fast' as const, + recommended: false, + repositoryUrl: + 'https://modelscope.cn/models/PaddlePaddle/PP-OCRv6_tiny_rec_onnx', + license: { + name: 'Apache License 2.0', + notice: '使用前请阅读模型许可证。', + url: 'https://example.com/license' + }, + files: [ + { + name: 'detection.onnx', + role: 'detection' as const, + download: { + url: 'https://modelscope.cn/models/example/detection.onnx', + size: 1_000, + sha256: 'a'.repeat(64) + } + }, + { + name: 'recognition.onnx', + role: 'recognition' as const, + download: { + url: 'https://modelscope.cn/models/example/recognition.onnx', + size: 2_000, + sha256: 'b'.repeat(64) + } + }, + { + name: 'dictionary.yml', + role: 'dictionary' as const, + download: { + url: 'https://modelscope.cn/models/example/dictionary.yml', + size: 500, + sha256: 'c'.repeat(64) + } + } + ] +} +const secondModelEntry = { + ...modelEntry, + id: 'pp-ocrv6-small', + displayName: 'PP-OCRv6 Small', + quality: 'balanced' as const, + speed: 'balanced' as const, + recommended: true +} +const thirdModelEntry = { + ...modelEntry, + id: 'pp-ocrv6-medium', + displayName: 'PP-OCRv6 Medium', + quality: 'high' as const, + speed: 'slow' as const, + recommended: false +} + +const snapshot: DocumentParsingSnapshot = { + settings, + status: { + nativeParsingAvailable: true, + conversionAvailable: false, + localOcr: { + id: 'pp-ocrv6-tiny', + displayName: 'PP-OCRv6 Tiny', + available: false, + verified: false, + runtime: 'onnxruntime-web-wasm', + detail: '模型尚未安装' + } + }, + ocrModels: { + rootDirectory: 'C:\\Users\\test\\models\\document-ocr', + catalog: [modelEntry, secondModelEntry, thirdModelEntry], + installed: [ + { + id: 'pp-ocrv6-small', + displayName: 'PP-OCRv6 Small', + source: 'download', + installedAt: '2026-08-11T00:00:00.000Z', + files: secondModelEntry.files.map((file) => ({ + name: file.name, + role: file.role, + size: file.download.size, + sha256: file.download.sha256 + })) + } + ], + operations: [] + } +} + +const getSnapshot = vi.fn(async () => snapshot) +const update = vi.fn(async (input: DocumentParsingSettings) => ({ + ...snapshot, + settings: input +})) +const test = vi.fn(async () => ({ + fileName: 'scan.pdf', + sourceFormat: 'PDF', + pageCount: 2, + ocrPageCount: 2, + characterCount: 120, + method: 'ocr' as const, + durationMs: 1_250, + preview: '扫描件识别正文', + warnings: [] +})) +const installOcrModel = vi.fn(async () => ({ + ...snapshot, + status: { + ...snapshot.status, + localOcr: { + ...snapshot.status.localOcr, + available: true, + verified: true, + detail: '模型已安装并校验' + } + }, + ocrModels: { + ...snapshot.ocrModels, + installed: [ + { + id: 'pp-ocrv6-tiny' as const, + displayName: 'PP-OCRv6 Tiny', + source: 'download' as const, + installedAt: '2026-08-11T00:00:00.000Z', + files: modelEntry.files.map((file) => ({ + name: file.name, + role: file.role, + size: file.download.size, + sha256: file.download.sha256 + })) + } + ] + } +})) +const importOcrModelArchive = vi.fn(async () => snapshot) +const exportOcrModelArchive = vi.fn(async () => snapshot) +const openOcrModelRepository = vi.fn(async () => undefined) + +describe('DocumentParsingSettingsSection', () => { + beforeEach(async () => { + await changeUiLocale('zh-CN') + vi.clearAllMocks() + Object.defineProperty(window, 'goodbuddy', { + configurable: true, + value: { + documentParsing: { + getSnapshot, + update, + test, + installOcrModel, + cancelOcrModelOperation: vi.fn(async () => true), + removeOcrModel: vi.fn(async () => snapshot), + importOcrModelArchive, + exportOcrModelArchive, + openOcrModelRepository, + openOcrModelsDirectory: vi.fn(), + getOcrAssets: vi.fn(), + respondOcr: vi.fn(), + onOcrRequest: vi.fn(() => () => undefined), + onOcrCancel: vi.fn(() => () => undefined) + } + } + }) + }) + + afterEach(() => cleanup()) + + it('shows actual capability status and saves workflow settings', async () => { + const onNotify = vi.fn() + render( + + ) + + expect(await screen.findByText('PP-OCRv6 Tiny')).toBeInTheDocument() + expect(screen.getByText('ModelScope')).toBeInTheDocument() + expect(screen.getByText('质量：基础')).toBeInTheDocument() + expect(screen.getByText('速度：快')).toBeInTheDocument() + expect(screen.getByText('旧版 Office 转换')).toBeInTheDocument() + expect( + screen.getByRole('button', { name: '本地模型' }) + ).toHaveAttribute('aria-pressed', 'true') + expect( + screen.getByRole('button', { + name: '远程服务（即将支持）' + }) + ).toBeDisabled() + expect(screen.queryByText('隐私与云端处理')).not.toBeInTheDocument() + expect( + screen.queryByText('模型详情与手动导入') + ).not.toBeInTheDocument() + fireEvent.click( + screen.getByRole('button', { + name: '打开 PP-OCRv6 Tiny 的 ModelScope 页面' + }) + ) + expect(openOcrModelRepository).toHaveBeenCalledWith('pp-ocrv6-tiny') + + fireEvent.change(screen.getByLabelText('聊天附件'), { + target: { value: 'fast-text' } + }) + fireEvent.click( + screen.getByRole('button', { name: '保存设置' }) + ) + + await waitFor(() => + expect(update).toHaveBeenCalledWith( + expect.objectContaining({ chatWorkflow: 'fast-text' }) + ) + ) + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: '文档解析设置已保存' + }) + ) + }) + + it('downloads the verified OCR model from the model catalog', async () => { + const onNotify = vi.fn() + render( + + ) + + fireEvent.click( + await screen.findByRole('button', { + name: '下载 PP-OCRv6 Tiny' + }) + ) + + await waitFor(() => + expect(installOcrModel).toHaveBeenCalledWith('pp-ocrv6-tiny') + ) + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: 'PP-OCRv6 Tiny 已安装' + }) + ) + }) + + it('imports and exports verified OCR model ZIP archives', async () => { + const onNotify = vi.fn() + render( + + ) + + fireEvent.click( + await screen.findByRole('button', { + name: '从 ZIP 导入 PP-OCRv6 Tiny' + }) + ) + await waitFor(() => + expect(importOcrModelArchive).toHaveBeenCalledWith( + 'pp-ocrv6-tiny' + ) + ) + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: 'PP-OCRv6 Tiny 已从 ZIP 导入' + }) + ) + + fireEvent.change(screen.getByLabelText('当前 OCR 模型'), { + target: { value: 'pp-ocrv6-small' } + }) + fireEvent.click( + screen.getByRole('button', { + name: '将 PP-OCRv6 Small 导出为 ZIP' + }) + ) + await waitFor(() => + expect(exportOcrModelArchive).toHaveBeenCalledWith( + 'pp-ocrv6-small' + ) + ) + expect(update).not.toHaveBeenCalled() + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: 'PP-OCRv6 Small 已导出为 ZIP' + }) + ) + }) + + it('switches the selected OCR model only when settings are saved', async () => { + render() + const selector = await screen.findByLabelText('当前 OCR 模型') + + expect( + screen.getByRole('option', { + name: 'PP-OCRv6 Tiny · 可下载' + }) + ).toBeInTheDocument() + expect( + screen.getByRole('option', { + name: 'PP-OCRv6 Small · 已安装' + }) + ).toBeInTheDocument() + expect( + screen.getByRole('option', { + name: 'PP-OCRv6 Medium · 可下载' + }) + ).toBeInTheDocument() + + fireEvent.change(selector, { + target: { value: 'pp-ocrv6-small' } + }) + + expect( + screen.getByText('模型选择尚未生效，点击“保存设置”后切换。') + ).toBeInTheDocument() + expect(screen.getByText('PP-OCRv6 Small')).toBeInTheDocument() + expect(screen.getByText('质量：均衡')).toBeInTheDocument() + expect(screen.getByText('速度：均衡')).toBeInTheDocument() + expect(update).not.toHaveBeenCalled() + + fireEvent.click( + screen.getByRole('button', { name: '保存设置' }) + ) + await waitFor(() => + expect(update).toHaveBeenCalledWith( + expect.objectContaining({ + localOcrModelId: 'pp-ocrv6-small' + }) + ) + ) + }) + + it('runs a real-file diagnostic flow and displays its result', async () => { + render() + await screen.findByText('PP-OCRv6 Tiny') + + fireEvent.click( + screen.getByRole('button', { name: '测试解析' }) + ) + + expect( + await screen.findByRole('dialog', { + name: '解析测试结果' + }) + ).toHaveTextContent('扫描件识别正文') + expect(test).toHaveBeenCalledOnce() + }) +}) diff --git a/src/renderer/src/DocumentParsingSettingsSection.tsx b/src/renderer/src/DocumentParsingSettingsSection.tsx new file mode 100644 index 0000000..f4af264 --- /dev/null +++ b/src/renderer/src/DocumentParsingSettingsSection.tsx @@ -0,0 +1,1017 @@ +import { + CheckCircle2, + Download, + ExternalLink, + FileSearch, + FileText, + FolderOpen, + ScanText, + ShieldCheck, + Square, + Trash2, + TriangleAlert, + Upload, + X +} from 'lucide-react' +import { + useCallback, + useEffect, + useRef, + useState +} from 'react' +import { useTranslation } from 'react-i18next' +import type { + DocumentParsingDiagnostic, + DocumentOcrModelCatalogEntry, + DocumentOcrModelOperation, + DocumentParsingSettings, + DocumentParsingSnapshot +} from '../../shared/document-parsing-contracts' +import type { AppNotificationInput } from './notifications' +import { SettingsCategoryHeader } from './SettingsPrimitives' +import { SegmentedControl } from './WorkspacePrimitives' + +type DocumentParsingSettingsSectionProps = { + onNotify?: (notification: AppNotificationInput) => void +} + +function errorMessage(reason: unknown, fallback: string): string { + if (!(reason instanceof Error)) { + return fallback + } + return reason.message.replace( + /^Error invoking remote method '[^']+': (?:Error: )?/u, + '' + ) +} + +function formatBytes(bytes: number): string { + return bytes >= 1024 * 1024 + ? `${(bytes / (1024 * 1024)).toFixed(1)} MB` + : `${(bytes / 1024).toFixed(1)} KB` +} + +function catalogSize(entry: DocumentOcrModelCatalogEntry): number { + return entry.files.reduce( + (total, file) => total + file.download.size, + 0 + ) +} + +function progressPercent( + operation: DocumentOcrModelOperation +): number | undefined { + return operation.totalBytes && operation.totalBytes > 0 + ? Math.min( + 100, + (operation.completedBytes / operation.totalBytes) * 100 + ) + : undefined +} + +function StatusRow({ + available, + detail, + label +}: { + available: boolean + detail: string + label: string +}): React.JSX.Element { + const { t } = useTranslation('settings') + return ( +
+ {available ? ( +
+ ) +} + +function DiagnosticDialog({ + diagnostic, + onClose +}: { + diagnostic: DocumentParsingDiagnostic + onClose: () => void +}): React.JSX.Element { + const { t } = useTranslation('settings') + const closeRef = useRef(null) + useEffect(() => { + closeRef.current?.focus() + }, []) + return ( +
{ + if (event.currentTarget === event.target) { + onClose() + } + }} + > +
{ + if (event.key === 'Escape') { + onClose() + } + }} + role="dialog" + > +
+
+ + {t('documentParsing.diagnostic.title')} + + {diagnostic.fileName} +
+ +
+
+
+
{t('documentParsing.diagnostic.format')}
+
{diagnostic.sourceFormat}
+
+
+
{t('documentParsing.diagnostic.method')}
+
+ {t( + `documentParsing.diagnostic.methods.${diagnostic.method}` + )} +
+
+
+
{t('documentParsing.diagnostic.pages')}
+
{diagnostic.pageCount}
+
+
+
{t('documentParsing.diagnostic.ocrPages')}
+
{diagnostic.ocrPageCount}
+
+
+
{t('documentParsing.diagnostic.characters')}
+
{diagnostic.characterCount.toLocaleString()}
+
+
+
{t('documentParsing.diagnostic.duration')}
+
{(diagnostic.durationMs / 1_000).toFixed(1)}s
+
+
+
+ {t('documentParsing.diagnostic.preview')} + 
{diagnostic.preview}
+
+ {diagnostic.warnings.length > 0 && ( +
+ {t('documentParsing.diagnostic.warnings')} +
    + {diagnostic.warnings.map((warning) => ( +
  • {warning}
    • + ))} +
+
+ )} +
+
+ ) +} + +export function DocumentParsingSettingsSection({ + onNotify +}: DocumentParsingSettingsSectionProps): React.JSX.Element { + const { t } = useTranslation('settings') + const [snapshot, setSnapshot] = useState() + const [draft, setDraft] = useState() + const [error, setError] = useState() + const [saving, setSaving] = useState(false) + const [testing, setTesting] = useState(false) + const [busyModelId, setBusyModelId] = useState() + const [confirmingRemove, setConfirmingRemove] = useState() + const [diagnostic, setDiagnostic] = + useState() + const mountedRef = useRef(false) + + const refresh = useCallback(async (): Promise => { + const api = window.goodbuddy.documentParsing + if (!api) { + throw new Error(t('errors.documentParsingUnavailable')) + } + const next = await api.getSnapshot() + if (mountedRef.current) { + setSnapshot(next) + } + }, [t]) + + useEffect(() => { + const api = window.goodbuddy.documentParsing + let active = true + mountedRef.current = true + if (!api) { + return () => { + active = false + mountedRef.current = false + } + } + void api.getSnapshot().then( + (next) => { + if (active) { + setSnapshot(next) + setDraft(next.settings) + } + }, + (reason: unknown) => { + if (active) { + setError( + errorMessage(reason, t('errors.readDocumentParsing')) + ) + } + } + ) + return () => { + active = false + mountedRef.current = false + } + }, [t]) + + const shouldPoll = + busyModelId !== undefined || + Boolean(snapshot?.ocrModels.operations.length) + + useEffect(() => { + if (!shouldPoll) { + return + } + const timer = window.setInterval(() => { + void refresh().catch(() => undefined) + }, 300) + return () => window.clearInterval(timer) + }, [refresh, shouldPoll]) + + const updateDraft = ( + key: Key, + value: DocumentParsingSettings[Key] + ): void => { + setDraft((current) => + current ? { ...current, [key]: value } : current + ) + } + + const save = async ( + notify = true + ): Promise => { + const api = window.goodbuddy.documentParsing + if (!api || !draft) { + return undefined + } + setSaving(true) + setError(undefined) + try { + const next = await api.update(draft) + setSnapshot(next) + setDraft(next.settings) + if (notify) { + onNotify?.({ + tone: 'success', + message: t('notifications.documentParsingSaved'), + dedupeKey: 'document-parsing-saved' + }) + } + return next + } catch (reason) { + setError( + errorMessage(reason, t('errors.saveDocumentParsing')) + ) + return undefined + } finally { + setSaving(false) + } + } + + const runModelOperation = async ( + modelId: string, + operation: () => Promise, + successMessage: string + ): Promise => { + setBusyModelId(modelId) + setConfirmingRemove(undefined) + setError(undefined) + try { + const next = await operation() + if (next && mountedRef.current) { + setSnapshot(next) + onNotify?.({ + tone: 'success', + message: successMessage, + dedupeKey: `document-ocr-model-${modelId}` + }) + } + } catch (reason) { + if (mountedRef.current) { + setError( + errorMessage( + reason, + t('errors.manageDocumentOcrModel') + ) + ) + } + } finally { + if (mountedRef.current) { + setBusyModelId(undefined) + void refresh().catch(() => undefined) + } + } + } + + const removeModel = async (modelId: string): Promise => { + const api = window.goodbuddy.documentParsing + if (!api) { + return + } + if (confirmingRemove !== modelId) { + setConfirmingRemove(modelId) + return + } + await runModelOperation( + modelId, + () => api.removeOcrModel(modelId), + t('documentParsing.ocr.notifications.removed') + ) + } + + const testParsing = async (): Promise => { + const api = window.goodbuddy.documentParsing + if (!api) { + setError(t('errors.documentParsingUnavailable')) + return + } + setTesting(true) + setError(undefined) + try { + if (!(await save(false))) { + return + } + const result = await api.test() + if (result) { + setDiagnostic(result) + onNotify?.({ + tone: 'success', + message: t('notifications.documentParsingTestSucceeded'), + dedupeKey: 'document-parsing-test' + }) + } + } catch (reason) { + setError( + errorMessage(reason, t('errors.testDocumentParsing')) + ) + } finally { + setTesting(false) + } + } + + if (!snapshot || !draft) { + const unavailableError = window.goodbuddy.documentParsing + ? undefined + : t('errors.documentParsingUnavailable') + return ( + <> + + {!error && !unavailableError && ( +

Loading…

+ )} + + ) + } + + const model = snapshot.ocrModels.catalog.find( + (entry) => entry.id === draft.localOcrModelId + ) + const installedModel = snapshot.ocrModels.installed.find( + (entry) => entry.id === draft.localOcrModelId + ) + const modelOperation = snapshot.ocrModels.operations.find( + (operation) => operation.modelId === draft.localOcrModelId + ) + const modelProgress = modelOperation + ? progressPercent(modelOperation) + : undefined + const pendingModelSelection = + draft.localOcrModelId !== snapshot.settings.localOcrModelId + + return ( + <> + + + + + } + category="document-parsing" + error={error} + /> + +
+
+
+
+ + + +
+

+ {t('documentParsing.status.partialNotice')} +

+
+ +
+
+
+
+ + +
+
+ +
+
+
+ +
+
+ {t('documentParsing.ocr.provider.title')} + + {t('documentParsing.ocr.provider.description')} + +
+ { + if (value === 'local') { + updateDraft('ocrProvider', value) + } + }} + options={[ + { + value: 'local', + label: t('documentParsing.ocr.provider.local') + }, + { + value: 'remote', + label: t('documentParsing.ocr.provider.remote'), + disabled: true + } + ]} + value={draft.ocrProvider} + /> + + {t('documentParsing.ocr.provider.remoteDescription')} + +
+ + {draft.ocrProvider === 'local' && ( + <> + + +

+ {t('documentParsing.ocr.storagePrefix')}{' '} + {snapshot.ocrModels.rootDirectory} + {t('documentParsing.ocr.storageSuffix')} +

+ + {model ? ( +
+
+
+
+ {model.displayName} + {model.recommended && ( + + {t('documentParsing.ocr.recommended')} + + )} +
+

{model.description}

+
+ ModelScope + + {model.languages.join(' / ')} + + + {model.runtime} + + + {t('documentParsing.ocr.quality.label', { + value: t( + `documentParsing.ocr.quality.values.${model.quality}` + ) + })} + + + {t('documentParsing.ocr.speed.label', { + value: t( + `documentParsing.ocr.speed.values.${model.speed}` + ) + })} + + + {formatBytes(catalogSize(model))} + + + {model.license.name} + +
+
+ +
+ +
+ + {installedModel && ( + +
+ +
+ {modelOperation ? ( + + ) : installedModel ? ( + <> + + + + ) : ( + <> + + + + )} +
+ + {modelOperation && ( +
+ + + {modelOperation.currentFile ?? + t('documentParsing.ocr.operations.preparing')} + {modelProgress === undefined + ? '' + : ` · ${modelProgress.toFixed(0)}%`} + +
+ )} +
+ ) : ( +

+ {t('documentParsing.ocr.catalogUnavailable')} +

+ )} + +
+ + +
+ + )} +
+ +
+ {t('documentParsing.advanced.title')} +
+ + + +
+ {t('documentParsing.advanced.concurrencyHint')} +
+ + {diagnostic && ( + setDiagnostic(undefined)} + /> + )} + + ) +} diff --git a/src/renderer/src/SettingsPanel.test.tsx b/src/renderer/src/SettingsPanel.test.tsx index b3c5dbb..238ea3d 100644 --- a/src/renderer/src/SettingsPanel.test.tsx +++ b/src/renderer/src/SettingsPanel.test.tsx @@ -478,7 +478,8 @@ describe('SettingsPanel runtime files', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select: selectSpeechModel, - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() }, diff --git a/src/renderer/src/SettingsPanel.tsx b/src/renderer/src/SettingsPanel.tsx index fae4854..d31bfba 100644 --- a/src/renderer/src/SettingsPanel.tsx +++ b/src/renderer/src/SettingsPanel.tsx @@ -38,6 +38,7 @@ import { UpdateSettingsSection } from './UpdateSettingsSection' import { PlatformFeaturesSettingsSection } from './PlatformFeaturesSettingsSection' import { SpeechModelSettingsSection } from './SpeechModelSettingsSection' import { EmbeddingSettingsSection } from './EmbeddingSettingsSection' +import { DocumentParsingSettingsSection } from './DocumentParsingSettingsSection' import { PageHeader, SegmentedControl } from './WorkspacePrimitives' import { SettingsCategoryHeader } from './SettingsPrimitives' import { @@ -346,6 +347,7 @@ export function SettingsPanel({ activeTab === 'roles' const categoryRendersOwnHeader = activeTab === 'platform-features' || + activeTab === 'document-parsing' || activeTab === 'channels' || activeTab === 'skills' || activeTab === 'mcp' || @@ -2258,6 +2260,10 @@ export function SettingsPanel({ )} + {activeTab === 'document-parsing' && ( + + )} + {activeTab === 'security' && ( <>
diff --git a/src/renderer/src/SpeechModelSettingsSection.test.tsx b/src/renderer/src/SpeechModelSettingsSection.test.tsx index 03c6899..826d57a 100644 --- a/src/renderer/src/SpeechModelSettingsSection.test.tsx +++ b/src/renderer/src/SpeechModelSettingsSection.test.tsx @@ -75,7 +75,8 @@ describe('SpeechModelSettingsSection', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select: vi.fn(), - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } @@ -128,7 +129,8 @@ describe('SpeechModelSettingsSection', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select: vi.fn(), - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } @@ -181,7 +183,8 @@ describe('SpeechModelSettingsSection', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select: vi.fn(), - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } @@ -200,6 +203,86 @@ describe('SpeechModelSettingsSection', () => { ) }) + it('imports and exports verified speech model ZIP archives', async () => { + const installedSnapshot: SpeechModelSnapshot = { + ...snapshot, + installed: [ + { + id: entry.id, + displayName: entry.displayName, + source: 'local', + installedAt: '2026-08-11T00:00:00.000Z', + files: [ + { + name: 'model.int8.onnx', + role: 'model', + size: 1_000, + sha256: 'a'.repeat(64) + } + ] + } + ] + } + const importArchive = vi.fn(async () => installedSnapshot) + const exportArchive = vi.fn(async () => installedSnapshot) + const select = vi.fn() + const onNotify = vi.fn() + const getSnapshot = vi + .fn<() => Promise>() + .mockResolvedValueOnce(snapshot) + .mockResolvedValue(installedSnapshot) + Object.defineProperty(window, 'goodbuddy', { + configurable: true, + value: { + speechModels: { + getSnapshot, + install: vi.fn(), + cancel: vi.fn(async () => true), + remove: vi.fn(), + select, + importArchive, + exportArchive, + openRepository: vi.fn(), + openModelsDirectory: vi.fn() + } + } as unknown as DesktopApi + }) + + render() + fireEvent.click( + await screen.findByRole('button', { + name: '从 ZIP 导入 SenseVoiceSmall INT8' + }) + ) + await waitFor(() => + expect(importArchive).toHaveBeenCalledWith( + 'sensevoice-small-int8' + ) + ) + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: 'SenseVoiceSmall INT8 已从 ZIP 导入' + }) + ) + + fireEvent.click( + await screen.findByRole('button', { + name: '将 SenseVoiceSmall INT8 导出为 ZIP' + }) + ) + await waitFor(() => + expect(exportArchive).toHaveBeenCalledWith( + 'sensevoice-small-int8' + ) + ) + expect(select).not.toHaveBeenCalled() + expect(onNotify).toHaveBeenCalledWith( + expect.objectContaining({ + message: 'SenseVoiceSmall INT8 已导出为 ZIP' + }) + ) + }) + it('shows live progress and cancellation for an active download', async () => { const active: SpeechModelSnapshot = { ...snapshot, @@ -224,7 +307,8 @@ describe('SpeechModelSettingsSection', () => { cancel, remove: vi.fn(), select: vi.fn(), - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } @@ -290,7 +374,8 @@ describe('SpeechModelSettingsSection', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select: vi.fn(), - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } @@ -355,7 +440,8 @@ describe('SpeechModelSettingsSection', () => { cancel: vi.fn(async () => true), remove: vi.fn(), select, - importLocalDirectory: vi.fn(), + importArchive: vi.fn(), + exportArchive: vi.fn(), openRepository: vi.fn(), openModelsDirectory: vi.fn() } diff --git a/src/renderer/src/SpeechModelSettingsSection.tsx b/src/renderer/src/SpeechModelSettingsSection.tsx index 1dcaf9d..23bad2e 100644 --- a/src/renderer/src/SpeechModelSettingsSection.tsx +++ b/src/renderer/src/SpeechModelSettingsSection.tsx @@ -6,7 +6,8 @@ import { FolderOpen, Mic, Square, - Trash2 + Trash2, + Upload } from 'lucide-react' import type { TFunction } from 'i18next' import { useCallback, useEffect, useRef, useState } from 'react' @@ -408,24 +409,49 @@ export function SpeechModelSettingsSection({ {t('speech.actions.cancel')} ) : installed ? ( - + <> + + + ) : ( <> {!entry.manualOnly && ( @@ -455,7 +481,7 @@ export function SpeechModelSettingsSection({ )} )} diff --git a/src/renderer/src/WorkspacePrimitives.tsx b/src/renderer/src/WorkspacePrimitives.tsx index 6aca9db..e13a743 100644 --- a/src/renderer/src/WorkspacePrimitives.tsx +++ b/src/renderer/src/WorkspacePrimitives.tsx @@ -30,6 +30,7 @@ export type PageTab = { export type SegmentedOption = { value: T label: string + disabled?: boolean } function nextControlIndex( @@ -258,11 +259,11 @@ export function SegmentedControl({ ? 'segmented-control__option segmented-control__option--active' : 'segmented-control__option' } - disabled={disabled} + disabled={disabled || option.disabled} key={option.value} onClick={() => onChange(option.value)} onKeyDown={(event) => { - const nextIndex = nextControlIndex( + let nextIndex = nextControlIndex( event, index, options.length @@ -270,6 +271,25 @@ export function SegmentedControl({ if (nextIndex === undefined) { return } + const direction = + event.key === 'ArrowLeft' || + event.key === 'ArrowUp' || + event.key === 'End' + ? -1 + : 1 + for ( + let attempts = 0; + attempts < options.length && + options[nextIndex]?.disabled; + attempts += 1 + ) { + nextIndex = + (nextIndex + direction + options.length) % + options.length + } + if (options[nextIndex]?.disabled) { + return + } event.preventDefault() onChange(options[nextIndex]!.value) const controls = diff --git a/src/renderer/src/document-ocr-bridge.ts b/src/renderer/src/document-ocr-bridge.ts new file mode 100644 index 0000000..8c79900 --- /dev/null +++ b/src/renderer/src/document-ocr-bridge.ts @@ -0,0 +1,203 @@ +import type { + DocumentOcrFailure, + DocumentOcrRequest, + DocumentOcrResult +} from '../../shared/document-parsing-contracts' + +type WorkerOutput = + | { type: 'ready' } + | { type: 'result'; result: DocumentOcrResult } + | { type: 'error'; requestId?: string; error: string } + +type PendingWorkerRequest = { + resolve: (result: DocumentOcrResult) => void + reject: (error: Error) => void + timer: number +} + +let worker: Worker | undefined +let workerModelId: string | undefined +let workerReady: Promise | undefined +let resolveWorkerReady: (() => void) | undefined +let rejectWorkerReady: ((error: Error) => void) | undefined +const pending = new Map() +const cancelledRequestIds = new Set() +let activeRequestId: string | undefined +let requestQueue: Promise = Promise.resolve() + +function safeError(error: unknown): string { + return error instanceof Error + ? error.message.slice(0, 1_000) + : '本地 OCR 解析失败' +} + +function terminateWorker(error: Error): void { + worker?.terminate() + rejectWorkerReady?.(error) + worker = undefined + workerModelId = undefined + workerReady = undefined + resolveWorkerReady = undefined + rejectWorkerReady = undefined + for (const request of pending.values()) { + window.clearTimeout(request.timer) + request.reject(error) + } + pending.clear() +} + +async function ensureWorker(modelId: string): Promise { + if (worker && workerReady && workerModelId === modelId) { + await workerReady + return worker + } + if (worker) { + terminateWorker(new Error('本地 OCR 模型已切换')) + } + const api = window.goodbuddy.documentParsing + if (!api) { + throw new Error('文档解析服务不可用') + } + worker = new Worker( + new URL('./document-ocr-worker.ts', import.meta.url), + { type: 'module', name: 'goodbuddy-document-ocr' } + ) + workerReady = new Promise((resolve, reject) => { + resolveWorkerReady = resolve + rejectWorkerReady = reject + }) + worker.addEventListener( + 'message', + (event: MessageEvent) => { + const output = event.data + if (output.type === 'ready') { + resolveWorkerReady?.() + return + } + if (output.type === 'error' && !output.requestId) { + rejectWorkerReady?.(new Error(output.error)) + return + } + const requestId = + output.type === 'result' + ? output.result.requestId + : output.requestId + if (!requestId) { + return + } + const request = pending.get(requestId) + if (!request) { + return + } + window.clearTimeout(request.timer) + pending.delete(requestId) + if (output.type === 'result') { + request.resolve(output.result) + } else { + request.reject(new Error(output.error)) + } + } + ) + worker.addEventListener('error', (event) => { + terminateWorker( + new Error(event.message || '本地 OCR Worker 异常') + ) + }) + try { + const assets = await api.getOcrAssets(modelId) + if (assets.modelId !== modelId) { + throw new Error('本地 OCR 模型与请求不匹配') + } + workerModelId = modelId + worker.postMessage( + { type: 'initialize', assets }, + [ + assets.detection, + assets.recognition, + assets.dictionary + ] + ) + await workerReady + } catch (error) { + terminateWorker( + error instanceof Error ? error : new Error('本地 OCR 初始化失败') + ) + throw error + } + return worker +} + +async function recognize( + request: DocumentOcrRequest +): Promise { + const activeWorker = await ensureWorker(request.modelId) + if (cancelledRequestIds.has(request.requestId)) { + throw new Error('本地 OCR 解析已取消') + } + const pageCount = request.pageNumbers?.length ?? request.maximumPages + const timeoutMs = Math.min( + 10 * 60 * 1_000, + Math.max( + request.pageTimeoutSeconds * 1_000, + request.pageTimeoutSeconds * pageCount * 1_000 + ) + ) + return new Promise((resolve, reject) => { + const timer = window.setTimeout(() => { + pending.delete(request.requestId) + terminateWorker(new Error('本地 OCR 解析超时')) + reject(new Error('本地 OCR 解析超时')) + }, timeoutMs) + pending.set(request.requestId, { resolve, reject, timer }) + activeWorker.postMessage( + { type: 'recognize', request }, + [request.data] + ) + }) +} + +async function handleRequest(request: DocumentOcrRequest): Promise { + const api = window.goodbuddy.documentParsing + if (!api || cancelledRequestIds.has(request.requestId)) { + cancelledRequestIds.delete(request.requestId) + return + } + activeRequestId = request.requestId + try { + await api.respondOcr(await recognize(request)) + } catch (error) { + const failure: DocumentOcrFailure = { + requestId: request.requestId, + error: safeError(error) + } + await api.respondOcr(failure).catch(() => undefined) + } finally { + activeRequestId = undefined + cancelledRequestIds.delete(request.requestId) + } +} + +export function installDocumentOcrBridge(): () => void { + const api = window.goodbuddy.documentParsing + if (!api) { + return () => undefined + } + const removeRequestListener = api.onOcrRequest((request) => { + requestQueue = requestQueue + .then(() => handleRequest(request)) + .catch(() => undefined) + }) + const removeCancelListener = api.onOcrCancel((requestId) => { + cancelledRequestIds.add(requestId) + if (activeRequestId === requestId) { + terminateWorker(new Error('本地 OCR 解析已取消')) + } + }) + return () => { + removeRequestListener() + removeCancelListener() + cancelledRequestIds.clear() + activeRequestId = undefined + terminateWorker(new Error('本地 OCR 服务已关闭')) + } +} diff --git a/src/renderer/src/document-ocr-worker.ts b/src/renderer/src/document-ocr-worker.ts new file mode 100644 index 0000000..96e5f30 --- /dev/null +++ b/src/renderer/src/document-ocr-worker.ts @@ -0,0 +1,233 @@ +/// + +import * as ort from 'onnxruntime-web' +import wasmModuleUrl from '../../../node_modules/onnxruntime-web/dist/ort-wasm-simd-threaded.mjs?url' +import wasmBinaryUrl from '../../../node_modules/onnxruntime-web/dist/ort-wasm-simd-threaded.wasm?url' +import pdfWorkerUrl from 'pdfjs-dist/build/pdf.worker.min.mjs?url' +import type { PDFPageProxy } from 'pdfjs-dist/types/src/display/api' +import type { + DocumentOcrAssets, + DocumentOcrRequest, + DocumentOcrResult +} from '../../shared/document-parsing-contracts' + +type InitializeMessage = { + type: 'initialize' + assets: DocumentOcrAssets +} + +type RecognizeMessage = { + type: 'recognize' + request: DocumentOcrRequest +} + +type WorkerInput = InitializeMessage | RecognizeMessage + +type WorkerOutput = + | { type: 'ready' } + | { type: 'result'; result: DocumentOcrResult } + | { type: 'error'; requestId?: string; error: string } + +type OcrService = InstanceType< + typeof import('ppu-paddle-ocr/web').PaddleOcrService +> + +const worker = self as DedicatedWorkerGlobalScope +let service: OcrService | undefined + +function absoluteAssetUrl(value: string): string { + return new URL(value, worker.location.href).href +} + +ort.env.wasm.numThreads = 1 +ort.env.wasm.proxy = false +ort.env.wasm.wasmPaths = { + mjs: absoluteAssetUrl(wasmModuleUrl), + wasm: absoluteAssetUrl(wasmBinaryUrl) +} + +function safeWorkerError(error: unknown): string { + return error instanceof Error + ? error.message.slice(0, 1_000) + : '本地 OCR 识别失败' +} + +async function initialize(assets: DocumentOcrAssets): Promise { + await service?.destroy() + const { PaddleOcrService } = await import('ppu-paddle-ocr/web') + service = new PaddleOcrService({ + model: { + detection: assets.detection, + recognition: assets.recognition, + charactersDictionary: assets.dictionary + }, + session: { + executionProviders: ['wasm'], + graphOptimizationLevel: 'disabled' + }, + processing: { + engine: 'canvas-native' + }, + recognition: { + charactersDictionary: [], + minimumConfidence: 0.5, + strategy: 'per-line', + recBatchSize: 4 + }, + detection: { + maxSideLength: 1920 + } + }) + await service.initialize() +} + +async function recognizeImage( + data: ArrayBuffer, + locator: string +): Promise { + if (!service?.isInitialized()) { + throw new Error('本地 OCR 模型尚未初始化') + } + const result = await service.recognize(data) + const content = result.text.replace(/\n{3,}/gu, '\n\n').trim() + return content + ? { + locator, + content, + confidence: result.confidence + } + : undefined +} + +async function renderPdfPage( + page: PDFPageProxy +): Promise { + const baseViewport = page.getViewport({ scale: 1 }) + const scale = Math.min( + 2, + 2200 / Math.max(baseViewport.width, baseViewport.height, 1) + ) + const viewport = page.getViewport({ scale }) + const canvas = new OffscreenCanvas( + Math.max(1, Math.ceil(viewport.width)), + Math.max(1, Math.ceil(viewport.height)) + ) + const context = canvas.getContext('2d', { + alpha: false, + willReadFrequently: true + }) + if (!context) { + throw new Error('无法创建 PDF 页面渲染画布') + } + await page.render({ + canvas: canvas as unknown as HTMLCanvasElement, + canvasContext: context as unknown as CanvasRenderingContext2D, + viewport + }).promise + const blob = await canvas.convertToBlob({ + type: 'image/png' + }) + return blob.arrayBuffer() +} + +async function recognizePdf( + request: DocumentOcrRequest +): Promise { + const pdfjs = await import('pdfjs-dist') + pdfjs.GlobalWorkerOptions.workerSrc = pdfWorkerUrl + const loadingTask = pdfjs.getDocument({ + data: new Uint8Array(request.data) + }) + const document = await loadingTask.promise + const selectedPages = new Set( + request.pageNumbers ?? + Array.from( + { length: Math.min(document.numPages, request.maximumPages) }, + (_, index) => index + 1 + ) + ) + if (document.numPages > request.maximumPages) { + await loadingTask.destroy() + throw new Error( + `PDF 共 ${document.numPages} 页，超过 ${request.maximumPages} 页限制` + ) + } + const sections: DocumentOcrResult['sections'] = [] + const warnings: string[] = [] + try { + for ( + let pageNumber = 1; + pageNumber <= document.numPages; + pageNumber += 1 + ) { + if (!selectedPages.has(pageNumber)) { + continue + } + const page = await document.getPage(pageNumber) + try { + const section = await recognizeImage( + await renderPdfPage(page), + `第 ${pageNumber} 页` + ) + if (section) { + sections.push(section) + } else { + warnings.push(`第 ${pageNumber} 页未识别到文字`) + } + } finally { + page.cleanup() + } + } + } finally { + await loadingTask.destroy() + } + return { + requestId: request.requestId, + sections, + pageCount: document.numPages, + warnings + } +} + +async function recognize(request: DocumentOcrRequest): Promise { + if (request.mimeType === 'application/pdf') { + return recognizePdf(request) + } + const section = await recognizeImage(request.data, '图片') + return { + requestId: request.requestId, + sections: section ? [section] : [], + pageCount: 1, + warnings: section ? [] : ['图片中未识别到文字'] + } +} + +worker.addEventListener('message', (event: MessageEvent) => { + const input = event.data + if (input.type === 'initialize') { + void initialize(input.assets).then( + () => worker.postMessage({ type: 'ready' } satisfies WorkerOutput), + (error: unknown) => + worker.postMessage({ + type: 'error', + error: safeWorkerError(error) + } satisfies WorkerOutput) + ) + return + } + void recognize(input.request).then( + (result) => + worker.postMessage({ + type: 'result', + result + } satisfies WorkerOutput), + (error: unknown) => + worker.postMessage({ + type: 'error', + requestId: input.request.requestId, + error: safeWorkerError(error) + } satisfies WorkerOutput) + ) +}) + +export {} diff --git a/src/renderer/src/i18n/locales/en-US/settings.ts b/src/renderer/src/i18n/locales/en-US/settings.ts index 09d089c..1a239b7 100644 --- a/src/renderer/src/i18n/locales/en-US/settings.ts +++ b/src/renderer/src/i18n/locales/en-US/settings.ts @@ -26,6 +26,12 @@ export const settings = { navigationDescription: 'LLMs, embedding models, and credentials', description: 'LLMs, embedding models, and credentials' }, + documentParsing: { + label: 'Document parsing', + navigationDescription: 'Attachments, knowledge, and local OCR', + description: + 'Configure extraction, conversion, and OCR for chat attachments and knowledge imports' + }, runtime: { label: 'Agent Runtime', navigationDescription: 'OpenCode, Continue, and workspace settings', @@ -76,6 +82,8 @@ export const settings = { saveAndTestRuntime: 'Save and test {{runtime}}', saving: 'Saving…', saveSettings: 'Save settings', + testParsing: 'Test parsing', + testingParsing: 'Parsing…', select: 'Select', selectFile: 'Select file', clear: 'Clear', @@ -115,11 +123,18 @@ export const settings = { openRuntimeConfig: 'Could not open the Runtime configuration', selectWorkspace: 'Could not select the workspace folder', retainModelConnection: 'Keep at least one model connection', - clearLocalData: 'Could not clear local data' + clearLocalData: 'Could not clear local data', + documentParsingUnavailable: 'Document parsing is unavailable', + readDocumentParsing: 'Could not load document parsing settings', + saveDocumentParsing: 'Could not save document parsing settings', + testDocumentParsing: 'Document parsing test failed', + manageDocumentOcrModel: 'OCR model operation failed' }, notifications: { settingsSaved: 'Settings saved', - connectionSucceeded: 'Connected: {{label}}' + connectionSucceeded: 'Connected: {{label}}', + documentParsingSaved: 'Document parsing settings saved', + documentParsingTestSucceeded: 'Document parsing test completed' }, credentials: { none: 'Not configured', @@ -212,6 +227,162 @@ export const settings = { 'Continue remains unavailable without a configuration file and will not load a remote default model anonymously.' } }, + documentParsing: { + status: { + title: 'Runtime status', + description: 'Capabilities currently available on this device', + available: 'Available', + unavailable: 'Unavailable', + verified: 'Verified', + native: 'Native document parsing', + nativeDetail: + 'Text, HTML, text PDFs, and modern Office documents', + conversion: 'Legacy Office conversion', + conversionUnavailable: + 'Not implemented yet; DOC, XLS, and PPT are currently unavailable', + localOcr: 'Local OCR', + ocrReady: + 'The model is installed, SHA-256 verified, and available offline', + ocrUnavailable: + 'The model is not installed or failed verification. Download it from ModelScope.', + partialNotice: + 'Basic document parsing is available. Legacy Office conversion is not implemented yet; scanned PDFs use local OCR.' + }, + workflows: { + title: 'Usage scenarios', + description: + 'Choose different parsing depth for chat attachments and knowledge imports', + chat: 'Chat attachments', + chatDescription: + 'Controls parsing before an attachment is added to the current request', + knowledge: 'Knowledge imports', + knowledgeDescription: + 'Controls parsing before chunking, indexing, and source location', + chatOptions: { + auto: 'Automatic parsing (recommended)', + fastText: 'Fast text', + highFidelity: 'High-fidelity parsing' + }, + knowledgeOptions: { + completeIndex: 'Complete indexing (recommended)', + fastIndex: 'Fast indexing', + highFidelity: 'High-fidelity indexing' + } + }, + ocr: { + title: 'OCR recognition', + description: + 'Install a local model on demand to recognize scanned PDFs on this device', + enabled: 'Enable local OCR', + enabledDescription: + 'After installation, the model runs only on this device through ONNX Runtime WebAssembly. Documents are not uploaded for recognition.', + model: 'Local model', + runtime: 'Runtime', + provider: { + title: 'OCR source', + description: + 'Choose either a local model or a remote service, then save settings to switch.', + local: 'Local model', + remote: 'Remote service (coming soon)', + remoteDescription: + 'Remote integrations will support services such as MinerU and PaddleOCR-VL. They are disabled in this version.' + }, + modelSelector: 'Current OCR model', + modelSelectorDescription: + 'This saved model is used for chat attachments and knowledge imports.', + pendingSelection: + 'This model selection is not active yet. Save settings to switch.', + installedOption: 'Installed', + downloadableOption: 'Available to download', + openModelsDirectory: 'Open model folder', + storagePrefix: 'Models are installed on demand in', + storageSuffix: + ' and can be exported as ZIP archives for offline devices.', + recommended: 'Recommended', + quality: { + label: 'Quality: {{value}}', + values: { + basic: 'Basic', + balanced: 'Balanced', + high: 'High' + } + }, + speed: { + label: 'Speed: {{value}}', + values: { + fast: 'Fast', + balanced: 'Balanced', + slow: 'Slow' + } + }, + installed: 'Installed and verified', + availableToDownload: 'Available from ModelScope', + download: 'Download', + importZip: 'Import ZIP', + exportZip: 'Export ZIP', + delete: 'Delete', + confirmDelete: 'Confirm delete', + cancel: 'Cancel', + openRepository: 'Open ModelScope', + catalogUnavailable: + 'No OCR model catalog is available in this version.', + mode: 'PDF OCR strategy', + modes: { + auto: 'Automatic; recognize only pages without useful text', + always: 'Always recognize every page', + disabled: 'Use only the PDF text layer' + }, + modelLicense: + 'The model uses Apache License 2.0 and is SHA-256 verified before loading.', + operations: { + preparing: 'Preparing model files', + downloading: 'Downloading from ModelScope', + importing: 'Importing model ZIP', + installing: 'Verifying and installing' + }, + accessibility: { + downloadModel: 'Download {{name}}', + importModelZip: 'Import {{name}} from a ZIP archive', + exportModelZip: 'Export {{name}} as a ZIP archive', + deleteModel: 'Delete {{name}}', + cancelOperation: 'Cancel {{name}} operation', + downloadProgress: '{{name}} download progress', + openRepository: 'Open the ModelScope page for {{name}}' + }, + notifications: { + installed: '{{name}} installed', + importedZip: '{{name}} imported from ZIP', + exportedZip: '{{name}} exported as ZIP', + removed: 'OCR model deleted' + } + }, + advanced: { + title: 'Advanced parsing settings', + maximumPages: 'Maximum OCR pages per document', + concurrency: 'OCR concurrency', + timeout: 'OCR time budget per page (seconds)', + concurrencyHint: + 'The WASM baseline currently processes pages serially; this value is reserved for batching and hardware acceleration.' + }, + diagnostic: { + title: 'Parsing test result', + file: 'File', + format: 'Format', + method: 'Method', + pages: 'Pages', + ocrPages: 'OCR pages', + characters: 'Extracted characters', + duration: 'Duration', + preview: 'Text preview', + warnings: 'Warnings', + methods: { + native: 'Native parsing', + ocr: 'Local OCR', + mixed: 'Native parsing and OCR' + }, + close: 'Close result' + } + }, model: { typeAriaLabel: 'Model type', types: { @@ -228,7 +399,7 @@ export const settings = { speech: { label: 'Speech model', description: - 'Select an installed model and save Settings to apply it; models can be downloaded or imported from a local folder.' + 'Select an installed model and save Settings to apply it; models can be downloaded or moved offline with ZIP archives.' } }, profile: { diff --git a/src/renderer/src/i18n/locales/en-US/settingsSections.ts b/src/renderer/src/i18n/locales/en-US/settingsSections.ts index 205ff28..53d45e7 100644 --- a/src/renderer/src/i18n/locales/en-US/settingsSections.ts +++ b/src/renderer/src/i18n/locales/en-US/settingsSections.ts @@ -7,11 +7,11 @@ export const settingsSections = { speech: { title: 'Speech models', description: - 'Model weights are not bundled. Download them as needed or import them from a local directory.', + 'Model weights are not bundled. Download them as needed or move them offline with ZIP archives.', openModelsDirectory: 'Open models directory', storagePrefix: 'Models are stored in', storageSuffix: - '. Automatic downloads pin the source revision and verify file sizes and SHA-256 hashes. You can also download models from their repositories and import them.', + '. Automatic downloads pin the source revision and verify SHA-256 hashes. Export a ZIP on an online device and import it directly on an offline device.', availableModels: 'Available speech models', loading: 'Loading speech models…', errors: { @@ -59,7 +59,8 @@ export const settingsSections = { delete: 'Delete', confirmDelete: 'Confirm delete', download: 'Download', - import: 'Import', + importZip: 'Import ZIP', + exportZip: 'Export ZIP', modelDetails: 'Model details', openRepository: 'Open model repository' }, @@ -69,13 +70,15 @@ export const settingsSections = { cancelOperation: 'Cancel the {{name}} operation', deleteModel: 'Delete {{name}}', downloadModel: 'Download {{name}}', - importModel: 'Import {{name}} from a local directory', + importModelZip: 'Import {{name}} from a ZIP archive', + exportModelZip: 'Export {{name}} as a ZIP archive', downloadProgress: '{{name}} download progress', openRepository: 'Open the {{name}} model repository' }, notifications: { installed: '{{name}} installed', - imported: '{{name}} imported from a local directory', + importedZip: '{{name}} imported from ZIP', + exportedZip: '{{name}} exported as ZIP', removed: 'Speech model deleted' }, details: { diff --git a/src/renderer/src/i18n/locales/zh-CN/settings.ts b/src/renderer/src/i18n/locales/zh-CN/settings.ts index 1840a24..2895b29 100644 --- a/src/renderer/src/i18n/locales/zh-CN/settings.ts +++ b/src/renderer/src/i18n/locales/zh-CN/settings.ts @@ -22,6 +22,11 @@ export const settings = { navigationDescription: 'LLM、向量模型与凭据', description: 'LLM、向量模型与凭据' }, + documentParsing: { + label: '文档解析', + navigationDescription: '附件、知识库与本地 OCR', + description: '统一配置聊天附件和知识库使用的提取、转换与 OCR 策略' + }, runtime: { label: 'Agent Runtime', navigationDescription: 'OpenCode、Continue 与工作区', @@ -69,6 +74,8 @@ export const settings = { saveAndTestRuntime: '保存并测试 {{runtime}}', saving: '保存中…', saveSettings: '保存设置', + testParsing: '测试解析', + testingParsing: '正在解析…', select: '选择', selectFile: '选择文件', clear: '清除', @@ -105,11 +112,18 @@ export const settings = { openRuntimeConfig: '打开 Runtime 配置失败', selectWorkspace: '选择工作区目录失败', retainModelConnection: '请至少保留一个模型连接', - clearLocalData: '本地数据清除失败' + clearLocalData: '本地数据清除失败', + documentParsingUnavailable: '文档解析服务不可用', + readDocumentParsing: '读取文档解析设置失败', + saveDocumentParsing: '保存文档解析设置失败', + testDocumentParsing: '测试文档解析失败', + manageDocumentOcrModel: 'OCR 模型操作失败' }, notifications: { settingsSaved: '设置已保存', - connectionSucceeded: '连接成功：{{label}}' + connectionSucceeded: '连接成功：{{label}}', + documentParsingSaved: '文档解析设置已保存', + documentParsingTestSucceeded: '文档解析测试完成' }, credentials: { none: '尚未配置', @@ -195,6 +209,149 @@ export const settings = { '未指定配置文件时 Continue 将保持不可用，不会匿名加载远程默认模型。' } }, + documentParsing: { + status: { + title: '运行状态', + description: '显示当前设备实际可用的解析能力', + available: '可用', + unavailable: '不可用', + verified: '已校验', + native: '原生文档解析', + nativeDetail: '文本、HTML、文本型 PDF 和新式 Office 文档', + conversion: '旧版 Office 转换', + conversionUnavailable: '尚未实现，DOC、XLS、PPT 暂不可用', + localOcr: '本地 OCR', + ocrReady: '模型已安装并通过 SHA-256 校验，可离线使用', + ocrUnavailable: '模型尚未安装或校验失败，请从 ModelScope 下载', + partialNotice: + '基础文档解析可用。旧版 Office 转换尚未实现；扫描 PDF 使用本地 OCR。' + }, + workflows: { + title: '使用场景', + description: '为聊天附件和知识库选择不同的解析深度', + chat: '聊天附件', + chatDescription: '控制附件加入当前请求前的解析方式', + knowledge: '知识库导入', + knowledgeDescription: '控制文档分块、索引和来源定位前的解析方式', + chatOptions: { + auto: '自动解析（推荐）', + fastText: '快速文本', + highFidelity: '高保真解析' + }, + knowledgeOptions: { + completeIndex: '完整索引（推荐）', + fastIndex: '快速索引', + highFidelity: '高保真索引' + } + }, + ocr: { + title: 'OCR 识别', + description: '按需安装本地模型，在设备上识别扫描 PDF', + enabled: '启用本地 OCR', + enabledDescription: + '模型安装后仅在本机通过 ONNX Runtime WebAssembly 运行，识别时不会上传文档。', + model: '本地模型', + runtime: '运行时', + provider: { + title: 'OCR 来源', + description: '本地模型与远程服务二选一，切换后保存设置生效。', + local: '本地模型', + remote: '远程服务（即将支持）', + remoteDescription: + '远程服务将支持 MinerU、PaddleOCR-VL 等接口，当前版本暂不可选。' + }, + modelSelector: '当前 OCR 模型', + modelSelectorDescription: '选择已保存，聊天附件和知识库将使用此模型。', + pendingSelection: '模型选择尚未生效，点击“保存设置”后切换。', + installedOption: '已安装', + downloadableOption: '可下载', + openModelsDirectory: '打开模型目录', + storagePrefix: '模型按需安装到', + storageSuffix: '。可导出 ZIP，并在内网设备直接导入。', + recommended: '推荐', + quality: { + label: '质量：{{value}}', + values: { + basic: '基础', + balanced: '均衡', + high: '高' + } + }, + speed: { + label: '速度：{{value}}', + values: { + fast: '快', + balanced: '均衡', + slow: '慢' + } + }, + installed: '已安装并校验', + availableToDownload: '可从 ModelScope 下载', + download: '下载', + importZip: '导入 ZIP', + exportZip: '导出 ZIP', + delete: '删除', + confirmDelete: '确认删除', + cancel: '取消', + openRepository: '打开 ModelScope', + catalogUnavailable: '当前版本没有可用的 OCR 模型目录。', + mode: 'PDF OCR 策略', + modes: { + auto: '自动，仅识别无有效文本的页面', + always: '始终识别所有页面', + disabled: '仅使用 PDF 文本层' + }, + modelLicense: + '模型采用 Apache License 2.0，并在加载前校验 SHA-256。', + operations: { + preparing: '正在准备模型文件', + downloading: '正在从 ModelScope 下载', + importing: '正在导入模型 ZIP', + installing: '正在校验并安装' + }, + accessibility: { + downloadModel: '下载 {{name}}', + importModelZip: '从 ZIP 导入 {{name}}', + exportModelZip: '将 {{name}} 导出为 ZIP', + deleteModel: '删除 {{name}}', + cancelOperation: '取消 {{name}} 操作', + downloadProgress: '{{name}} 下载进度', + openRepository: '打开 {{name}} 的 ModelScope 页面' + }, + notifications: { + installed: '{{name}} 已安装', + importedZip: '{{name}} 已从 ZIP 导入', + exportedZip: '{{name}} 已导出为 ZIP', + removed: 'OCR 模型已删除' + } + }, + advanced: { + title: '高级解析设置', + maximumPages: '单文档最大 OCR 页数', + concurrency: 'OCR 并发数', + timeout: '每页 OCR 时间预算（秒）', + concurrencyHint: + '当前 WASM 基线按页串行执行；该值为后续批处理和硬件加速保留。' + }, + diagnostic: { + title: '解析测试结果', + file: '文件', + format: '格式', + method: '处理方式', + pages: '页数', + ocrPages: 'OCR 页数', + characters: '提取字符', + duration: '耗时', + preview: '文本预览', + warnings: '警告', + methods: { + native: '原生解析', + ocr: '本地 OCR', + mixed: '原生解析与 OCR' + }, + close: '关闭结果' + } + }, model: { typeAriaLabel: '模型类型', types: { @@ -209,7 +366,7 @@ export const settings = { speech: { label: '语音模型', description: - '选择已安装模型后保存设置生效；模型可按需下载或从本地目录导入。' + '选择已安装模型后保存设置生效；模型可按需下载或通过 ZIP 离线迁移。' } }, profile: { diff --git a/src/renderer/src/i18n/locales/zh-CN/settingsSections.ts b/src/renderer/src/i18n/locales/zh-CN/settingsSections.ts index 99374fc..c58daf8 100644 --- a/src/renderer/src/i18n/locales/zh-CN/settingsSections.ts +++ b/src/renderer/src/i18n/locales/zh-CN/settingsSections.ts @@ -1,11 +1,11 @@ export const settingsSections = { speech: { title: '语音模型', - description: '应用不内置模型权重，按需下载或从本地目录导入', + description: '应用不内置模型权重，按需下载或通过 ZIP 离线迁移', openModelsDirectory: '打开模型目录', storagePrefix: '模型保存在', storageSuffix: - '。自动下载会固定来源版本，并校验文件大小和 SHA-256；也可以从模型仓库手动下载后导入。', + '。自动下载会固定来源版本并校验 SHA-256；外网设备可导出 ZIP，内网设备可直接导入。', availableModels: '可用语音模型', loading: '正在读取语音模型…', errors: { @@ -52,7 +52,8 @@ export const settingsSections = { delete: '删除', confirmDelete: '确认删除', download: '下载', - import: '导入', + importZip: '导入 ZIP', + exportZip: '导出 ZIP', modelDetails: '模型详情', openRepository: '打开模型仓库' }, @@ -62,13 +63,15 @@ export const settingsSections = { cancelOperation: '取消 {{name}} 操作', deleteModel: '删除 {{name}}', downloadModel: '下载 {{name}}', - importModel: '从本地目录导入 {{name}}', + importModelZip: '从 ZIP 导入 {{name}}', + exportModelZip: '将 {{name}} 导出为 ZIP', downloadProgress: '{{name}}下载进度', openRepository: '打开 {{name}} 模型仓库' }, notifications: { installed: '{{name}} 已安装', - imported: '{{name}} 已从本地目录导入', + importedZip: '{{name}} 已从 ZIP 导入', + exportedZip: '{{name}} 已导出为 ZIP', removed: '语音模型已删除' }, details: { diff --git a/src/renderer/src/main.tsx b/src/renderer/src/main.tsx index 4340111..69ada46 100644 --- a/src/renderer/src/main.tsx +++ b/src/renderer/src/main.tsx @@ -17,6 +17,7 @@ import { loadAppearanceTheme, resolveAppearanceTheme } from './theme' +import { installDocumentOcrBridge } from './document-ocr-bridge' import './styles.css' const root = document.getElementById('root') @@ -43,6 +44,8 @@ applyAppearanceTheme( ) ) +installDocumentOcrBridge() + createRoot(root).render( :not(summary) + :not(summary) { } } +.document-parsing-status__list { + display: grid; + grid-template-columns: repeat(3, minmax(0, 1fr)); + gap: var(--space-3); +} + +.document-parsing-status__row { + display: grid; + min-width: 0; + align-items: center; + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); + grid-template-columns: auto minmax(0, 1fr); + gap: var(--space-2) var(--space-3); +} + +.document-parsing-status__row > svg { + color: var(--text-muted); +} + +.document-parsing-status__row:has( + .document-parsing-status__badge--available + ) + > svg { + color: var(--success); +} + +.document-parsing-status__row > span:nth-child(2), +.document-parsing-model > span:first-child { + display: grid; + min-width: 0; + gap: var(--space-1); +} + +.document-parsing-status__row strong, +.document-parsing-model strong { + color: var(--text-primary); + font-size: var(--font-body); +} + +.document-parsing-status__row small, +.document-parsing-model small { + color: var(--text-muted); + font-size: var(--font-caption); + line-height: 1.5; +} + +.document-parsing-status__badge { + width: fit-content; + padding: var(--space-1) var(--space-2); + border-radius: var(--radius-control); + background: var(--surface-muted); + color: var(--text-muted); + font-size: var(--font-caption); + grid-column: 2; + white-space: nowrap; +} + +.document-parsing-status__badge--available { + background: var(--success-subtle); + color: var(--success); +} + +.document-parsing-grid { + display: grid; + grid-template-columns: repeat(2, minmax(0, 1fr)); + gap: var(--space-4); +} + +.document-parsing-grid .field > small { + color: var(--text-muted); + font-size: var(--font-caption); + line-height: 1.5; +} + +.document-parsing-workflows .field { + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); +} + +.document-ocr-settings .settings-section__title--actions > button, +.document-ocr-model__actions, +.document-ocr-model__actions button, +.document-ocr-model__repository { + display: flex; + align-items: center; +} + +.document-ocr-settings .settings-section__title--actions > button, +.document-ocr-model__actions button, +.document-ocr-model__repository { + gap: var(--space-2); +} + +.document-ocr-settings__storage { + overflow: hidden; +} + +.document-ocr-settings__storage code { + overflow-wrap: anywhere; +} + +.document-ocr-provider { + display: grid; + align-items: center; + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); + grid-template-columns: minmax(0, 1fr) auto; + gap: var(--space-2) var(--space-4); +} + +.document-ocr-provider > div:first-child { + display: grid; + gap: var(--space-1); +} + +.document-ocr-provider strong { + color: var(--text-primary); + font-size: var(--font-body); +} + +.document-ocr-provider small { + color: var(--text-muted); + font-size: var(--font-caption); + line-height: 1.5; +} + +.document-ocr-provider > small { + grid-column: 1 / -1; +} + +.document-ocr-model-selector { + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); +} + +.document-ocr-model-selector > small { + color: var(--text-muted); + font-size: var(--font-caption); + line-height: 1.5; +} + +.document-ocr-model { + display: grid; + min-width: 0; + align-items: start; + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); + grid-template-columns: minmax(0, 1fr) minmax(150px, auto); + gap: var(--space-2) var(--space-4); +} + +.document-ocr-model__summary { + display: grid; + min-width: 0; + gap: var(--space-1); +} + +.document-ocr-model__header { + display: flex; + min-width: 0; + align-items: flex-start; + justify-content: space-between; + grid-column: 1 / -1; + gap: var(--space-4); +} + +.document-ocr-model__repository { + min-height: 30px; + flex: 0 0 auto; + white-space: nowrap; +} + +.document-ocr-model__name, +.document-ocr-model__tags, +.document-ocr-model__status { + display: flex; + align-items: center; +} + +.document-ocr-model__name, +.document-ocr-model__tags { + flex-wrap: wrap; + gap: var(--space-2); +} + +.document-ocr-model__summary strong { + color: var(--text-primary); + font-size: var(--font-body); +} + +.document-ocr-model__summary p { + margin: 0; + color: var(--text-secondary); + font-size: var(--font-caption); + line-height: 1.55; +} + +.document-ocr-model__state { + display: flex; + justify-content: flex-start; +} + +.document-ocr-model__status { + color: var(--text-muted); + font-size: var(--font-caption); + font-weight: 650; + gap: var(--space-1); + white-space: nowrap; +} + +.document-ocr-model__status--installed { + color: var(--success); +} + +.document-ocr-model__actions { + justify-content: flex-end; + flex-wrap: wrap; + grid-column: 2; + gap: var(--space-2); +} + +.document-ocr-model__actions button { + min-height: 30px; + white-space: nowrap; +} + +.document-ocr-model__actions .danger-ghost { + display: flex; + align-items: center; + padding: 0 var(--space-2); + border: 1px solid transparent; + border-radius: var(--radius-control); + background: transparent; + color: var(--danger); + font: inherit; + font-size: var(--font-caption); + gap: var(--space-1); +} + +.document-ocr-model__actions .danger-ghost:hover { + border-color: var(--danger-border); + background: var(--danger-subtle); +} + +.document-ocr-model__operation { + display: grid; + grid-column: 1 / -1; + gap: var(--space-1); +} + +.document-ocr-model__operation progress { + width: 100%; + accent-color: var(--accent-solid); +} + +.document-ocr-model__operation small { + color: var(--text-muted); + font-size: var(--font-caption); +} + +.document-ocr-settings__options { + display: grid; + align-items: start; + grid-template-columns: repeat(2, minmax(0, 1fr)); + gap: var(--space-4); +} + +.document-ocr-settings__options > * { + min-height: 100%; + padding: var(--space-3); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); +} + +.document-ocr-settings__options .settings-checkbox { + display: grid; + align-items: start; + grid-template-columns: auto minmax(0, 1fr); + gap: var(--space-3); +} + +.document-ocr-settings__options .settings-checkbox > input { + width: 16px; + height: 16px; + margin: 2px 0 0; + accent-color: var(--accent-solid); +} + +.document-ocr-settings__options .settings-checkbox > span { + display: grid; + gap: var(--space-1); +} + +.document-ocr-settings__options .settings-checkbox strong { + color: var(--text-primary); + font-size: var(--font-body); +} + +.document-ocr-settings__options .settings-checkbox small { + color: var(--text-muted); + font-size: var(--font-caption); + line-height: 1.5; +} + +.document-parsing-diagnostic-backdrop { + position: fixed; + z-index: var(--z-dialog); + display: grid; + padding: var(--space-4); + background: rgb(5 12 24 / 58%); + inset: 0; + place-items: center; +} + +.document-parsing-diagnostic { + display: grid; + width: min(620px, 100%); + max-height: min(760px, calc(100vh - 32px)); + overflow-y: auto; + padding: var(--space-4); + border: 1px solid var(--border-default); + border-radius: var(--radius-card); + background: var(--surface-raised); + box-shadow: var(--shadow-dialog); + gap: var(--space-4); +} + +.document-parsing-diagnostic > header { + display: flex; + align-items: flex-start; + justify-content: space-between; + gap: var(--space-4); +} + +.document-parsing-diagnostic > header > div { + display: grid; + gap: var(--space-1); +} + +.document-parsing-diagnostic > header strong { + color: var(--text-primary); + font-size: var(--font-section-title); +} + +.document-parsing-diagnostic > header small { + color: var(--text-muted); + font-size: var(--font-caption); +} + +.document-parsing-diagnostic dl { + display: grid; + margin: 0; + grid-template-columns: repeat(3, minmax(0, 1fr)); + gap: var(--space-3); +} + +.document-parsing-diagnostic dl > div { + display: grid; + padding: var(--space-3); + border-radius: var(--radius-control); + background: var(--surface-subtle); + gap: var(--space-1); +} + +.document-parsing-diagnostic dt { + color: var(--text-muted); + font-size: var(--font-caption); +} + +.document-parsing-diagnostic dd { + margin: 0; + color: var(--text-primary); + font-size: var(--font-body); + font-weight: 650; +} + +.document-parsing-diagnostic__preview { + display: grid; + min-width: 0; + gap: var(--space-2); +} + +.document-parsing-diagnostic__preview pre { + max-height: 260px; + overflow: auto; + margin: 0; + padding: var(--space-3); + border: 1px solid var(--border-subtle); + border-radius: var(--radius-control); + background: var(--surface-subtle); + color: var(--text-secondary); + font-family: var(--font-family-mono); + font-size: var(--font-caption); + line-height: 1.55; + white-space: pre-wrap; + word-break: break-word; +} + +@media (max-width: 720px) { + .document-parsing-status__list, + .document-parsing-grid, + .document-parsing-diagnostic dl, + .document-ocr-settings__options { + grid-template-columns: minmax(0, 1fr); + } + + .document-ocr-settings .settings-section__title--actions { + align-items: flex-start; + flex-wrap: wrap; + } + + .document-ocr-provider { + grid-template-columns: minmax(0, 1fr); + } + + .document-ocr-provider > small { + grid-column: 1; + } + + .document-ocr-model { + grid-template-columns: minmax(0, 1fr); + } + + .document-ocr-model__header { + align-items: stretch; + flex-direction: column; + } + + .document-ocr-model__repository { + width: fit-content; + } + + .document-ocr-model__state, + .document-ocr-model__actions { + justify-content: flex-start; + grid-column: 1; + } +} + .role-prompt-empty { min-height: 180px; padding: var(--space-6); diff --git a/src/shared/contracts.ts b/src/shared/contracts.ts index cd9508d..4c70ba5 100644 --- a/src/shared/contracts.ts +++ b/src/shared/contracts.ts @@ -74,6 +74,15 @@ import type { EmbeddingIndexStatus, EmbeddingSettingsSnapshot } from './embedding-contracts' +import type { + DocumentOcrAssets, + DocumentOcrFailure, + DocumentOcrRequest, + DocumentOcrResult, + DocumentParsingDiagnostic, + DocumentParsingSettings, + DocumentParsingSnapshot +} from './document-parsing-contracts' import type { WeixinBindingSnapshot } from './weixin-channel-contracts' import type { RemoteChannelActivity } from './remote-channel-contracts' import { @@ -1038,7 +1047,10 @@ export type DesktopApi = { cancel: (modelId: string) => Promise remove: (modelId: string) => Promise select: (modelId: string | null) => Promise - importLocalDirectory: ( + importArchive: ( + modelId: string + ) => Promise + exportArchive: ( modelId: string ) => Promise openRepository: (modelId: string) => Promise @@ -1059,6 +1071,38 @@ export type DesktopApi = { listener: (status: EmbeddingIndexStatus) => void ) => () => void } + documentParsing?: { + getSnapshot: () => Promise + update: ( + input: DocumentParsingSettings + ) => Promise + test: () => Promise + installOcrModel: ( + modelId: string + ) => Promise + cancelOcrModelOperation: (modelId: string) => Promise + removeOcrModel: ( + modelId: string + ) => Promise + importOcrModelArchive: ( + modelId: string + ) => Promise + exportOcrModelArchive: ( + modelId: string + ) => Promise + openOcrModelRepository: (modelId: string) => Promise + openOcrModelsDirectory: () => Promise + getOcrAssets: (modelId: string) => Promise + respondOcr: ( + response: DocumentOcrResult | DocumentOcrFailure + ) => Promise + onOcrRequest: ( + listener: (request: DocumentOcrRequest) => void + ) => () => void + onOcrCancel: ( + listener: (requestId: string) => void + ) => () => void + } projects: { list: (includeArchived?: boolean) => Promise create: (input: ProjectCreateInput) => Promise diff --git a/src/shared/document-parsing-contracts.ts b/src/shared/document-parsing-contracts.ts new file mode 100644 index 0000000..d2ee11d --- /dev/null +++ b/src/shared/document-parsing-contracts.ts @@ -0,0 +1,310 @@ +import { z } from 'zod' + +export const documentParsingPurposeSchema = z.enum([ + 'chat-attachment', + 'knowledge-index', + 'diagnostic' +]) + +export const chatDocumentWorkflowSchema = z.enum([ + 'auto', + 'fast-text', + 'high-fidelity' +]) + +export const knowledgeDocumentWorkflowSchema = z.enum([ + 'complete-index', + 'fast-index', + 'high-fidelity' +]) + +export const pdfOcrModeSchema = z.enum([ + 'auto', + 'always', + 'disabled' +]) + +export const documentOcrProviderSchema = z.literal('local') + +export const localOcrModelIdSchema = z + .string() + .min(1) + .max(96) + .regex(/^[a-z0-9]+(?:-[a-z0-9]+)*$/u) + +const documentOcrSha256Schema = z + .string() + .regex(/^[a-f0-9]{64}$/u) + +export const documentOcrModelFileRoleSchema = z.enum([ + 'detection', + 'recognition', + 'dictionary' +]) + +export const documentOcrModelDownloadSchema = z + .object({ + url: z.url().max(2_048), + size: z.number().int().positive().safe(), + sha256: documentOcrSha256Schema + }) + .strict() + +export const documentOcrModelFileSchema = z + .object({ + name: z + .string() + .min(1) + .max(255) + .regex(/^[^/\\\0]+$/u), + role: documentOcrModelFileRoleSchema, + download: documentOcrModelDownloadSchema + }) + .strict() + +export const documentOcrModelCatalogEntrySchema = z + .object({ + id: localOcrModelIdSchema, + displayName: z.string().trim().min(1).max(120), + description: z.string().trim().min(1).max(500), + languages: z.array(z.string().trim().min(1).max(32)).min(1).max(32), + runtime: z.literal('onnxruntime-web-wasm'), + quality: z.enum(['basic', 'balanced', 'high']), + speed: z.enum(['fast', 'balanced', 'slow']), + recommended: z.boolean(), + repositoryUrl: z.url().max(2_048), + license: z + .object({ + name: z.string().trim().min(1).max(120), + notice: z.string().trim().min(1).max(1_000), + url: z.url().max(2_048) + }) + .strict(), + files: z.array(documentOcrModelFileSchema).length(3) + }) + .strict() + .superRefine((entry, context) => { + if ( + new Set(entry.files.map((file) => file.name)).size !== + entry.files.length + ) { + context.addIssue({ + code: 'custom', + path: ['files'], + message: 'OCR 模型文件名不能重复' + }) + } + if ( + new Set(entry.files.map((file) => file.role)).size !== + entry.files.length + ) { + context.addIssue({ + code: 'custom', + path: ['files'], + message: 'OCR 模型文件角色不能重复' + }) + } + }) + +export const installedDocumentOcrModelSchema = z + .object({ + id: localOcrModelIdSchema, + displayName: z.string().trim().min(1).max(120), + source: z.enum(['download', 'local']), + installedAt: z.string().datetime(), + files: z + .array( + z + .object({ + name: z.string().min(1).max(255).regex(/^[^/\\\0]+$/u), + role: documentOcrModelFileRoleSchema, + size: z.number().int().positive().safe(), + sha256: documentOcrSha256Schema + }) + .strict() + ) + .length(3) + }) + .strict() + +export const documentOcrModelOperationSchema = z + .object({ + modelId: localOcrModelIdSchema, + kind: z.enum(['download', 'import']), + phase: z.enum(['preparing', 'transferring', 'installing']), + currentFile: z.string().min(1).max(255).nullable(), + completedBytes: z.number().int().nonnegative().safe(), + totalBytes: z.number().int().nonnegative().safe().nullable() + }) + .strict() + +export const documentOcrModelSnapshotSchema = z + .object({ + rootDirectory: z.string().min(1).max(32_768), + catalog: z.array(documentOcrModelCatalogEntrySchema).max(16), + installed: z.array(installedDocumentOcrModelSchema).max(16), + operations: z.array(documentOcrModelOperationSchema).max(8) + }) + .strict() + +export const documentOcrModelActionInputSchema = z + .object({ + modelId: localOcrModelIdSchema + }) + .strict() + +export const documentParsingSettingsSchema = z + .object({ + chatWorkflow: chatDocumentWorkflowSchema, + knowledgeWorkflow: knowledgeDocumentWorkflowSchema, + pdfOcrMode: pdfOcrModeSchema, + ocrProvider: documentOcrProviderSchema, + localOcrEnabled: z.boolean(), + localOcrModelId: localOcrModelIdSchema, + maximumPages: z.number().int().min(1).max(500), + ocrConcurrency: z.number().int().min(1).max(4), + pageTimeoutSeconds: z.number().int().min(10).max(300) + }) + .strict() + +export const documentParsingSettingsUpdateSchema = + documentParsingSettingsSchema + +export const documentParsingModelStatusSchema = z + .object({ + id: localOcrModelIdSchema, + displayName: z.string().trim().min(1).max(120), + available: z.boolean(), + verified: z.boolean(), + runtime: z.literal('onnxruntime-web-wasm'), + detail: z.string().trim().min(1).max(500) + }) + .strict() + +export const documentParsingStatusSchema = z + .object({ + nativeParsingAvailable: z.literal(true), + conversionAvailable: z.boolean(), + localOcr: documentParsingModelStatusSchema + }) + .strict() + +export const documentParsingSnapshotSchema = z + .object({ + settings: documentParsingSettingsSchema, + status: documentParsingStatusSchema, + ocrModels: documentOcrModelSnapshotSchema + }) + .strict() + +export const documentParsingTestInputSchema = z + .object({ + purpose: documentParsingPurposeSchema.default('diagnostic') + }) + .strict() + +export const documentParsingDiagnosticSchema = z + .object({ + fileName: z.string().trim().min(1).max(500), + sourceFormat: z.string().trim().min(1).max(32), + pageCount: z.number().int().nonnegative().max(10_000), + ocrPageCount: z.number().int().nonnegative().max(10_000), + characterCount: z.number().int().nonnegative().safe(), + method: z.enum(['native', 'ocr', 'mixed']), + durationMs: z.number().int().nonnegative().safe(), + preview: z.string().max(2_000), + warnings: z.array(z.string().trim().min(1).max(500)).max(20) + }) + .strict() + +export const documentOcrAssetsSchema = z + .object({ + modelId: localOcrModelIdSchema, + detection: z.instanceof(ArrayBuffer), + recognition: z.instanceof(ArrayBuffer), + dictionary: z.instanceof(ArrayBuffer) + }) + .strict() + +export const documentOcrRequestSchema = z + .object({ + requestId: z.string().uuid(), + modelId: localOcrModelIdSchema, + fileName: z.string().trim().min(1).max(500), + mimeType: z.enum([ + 'application/pdf', + 'image/jpeg', + 'image/png', + 'image/webp' + ]), + data: z + .instanceof(ArrayBuffer) + .refine( + (value) => value.byteLength > 0 && value.byteLength <= 20 * 1024 * 1024, + 'OCR 输入必须介于 1 字节和 20MB 之间' + ), + maximumPages: z.number().int().min(1).max(500), + pageNumbers: z + .array(z.number().int().min(1).max(10_000)) + .min(1) + .max(500) + .optional(), + pageTimeoutSeconds: z.number().int().min(10).max(300) + }) + .strict() + +export const documentOcrSectionSchema = z + .object({ + locator: z.string().trim().min(1).max(500), + content: z.string().trim().min(1).max(1_000_000), + confidence: z.number().min(0).max(1) + }) + .strict() + +export const documentOcrResultSchema = z + .object({ + requestId: z.string().uuid(), + sections: z.array(documentOcrSectionSchema).max(500), + pageCount: z.number().int().nonnegative().max(10_000), + warnings: z.array(z.string().trim().min(1).max(500)).max(20) + }) + .strict() + +export const documentOcrFailureSchema = z + .object({ + requestId: z.string().uuid(), + error: z.string().trim().min(1).max(1_000) + }) + .strict() + +export type DocumentParsingPurpose = z.infer< + typeof documentParsingPurposeSchema +> +export type DocumentParsingSettings = z.infer< + typeof documentParsingSettingsSchema +> +export type DocumentParsingSnapshot = z.infer< + typeof documentParsingSnapshotSchema +> +export type DocumentOcrModelFile = z.infer< + typeof documentOcrModelFileSchema +> +export type DocumentOcrModelCatalogEntry = z.infer< + typeof documentOcrModelCatalogEntrySchema +> +export type InstalledDocumentOcrModel = z.infer< + typeof installedDocumentOcrModelSchema +> +export type DocumentOcrModelOperation = z.infer< + typeof documentOcrModelOperationSchema +> +export type DocumentOcrModelSnapshot = z.infer< + typeof documentOcrModelSnapshotSchema +> +export type DocumentParsingDiagnostic = z.infer< + typeof documentParsingDiagnosticSchema +> +export type DocumentOcrAssets = z.infer +export type DocumentOcrRequest = z.infer +export type DocumentOcrResult = z.infer +export type DocumentOcrFailure = z.infer diff --git a/src/shared/ipc-channels.ts b/src/shared/ipc-channels.ts index b53be1e..b1c6897 100644 --- a/src/shared/ipc-channels.ts +++ b/src/shared/ipc-channels.ts @@ -48,7 +48,8 @@ export const ipcChannels = { speechModelsCancel: 'settings:speech-models:cancel', speechModelsRemove: 'settings:speech-models:remove', speechModelsSelect: 'settings:speech-models:select', - speechModelsImportLocal: 'settings:speech-models:import-local', + speechModelsImportArchive: 'settings:speech-models:import-archive', + speechModelsExportArchive: 'settings:speech-models:export-archive', speechModelsOpenRepository: 'settings:speech-models:open-repository', speechModelsOpenDirectory: 'settings:speech-models:open-directory', speechTranscribe: 'speech:transcribe', @@ -58,6 +59,24 @@ export const ipcChannels = { embeddingIndexRebuild: 'settings:embedding:index:rebuild', embeddingIndexCancel: 'settings:embedding:index:cancel', embeddingIndexStatusChanged: 'settings:embedding:index:status-changed', + documentParsingGet: 'settings:document-parsing:get', + documentParsingUpdate: 'settings:document-parsing:update', + documentParsingTest: 'settings:document-parsing:test', + documentOcrModelsInstall: 'settings:document-ocr-models:install', + documentOcrModelsCancel: 'settings:document-ocr-models:cancel', + documentOcrModelsRemove: 'settings:document-ocr-models:remove', + documentOcrModelsImportArchive: + 'settings:document-ocr-models:import-archive', + documentOcrModelsExportArchive: + 'settings:document-ocr-models:export-archive', + documentOcrModelsOpenRepository: + 'settings:document-ocr-models:open-repository', + documentOcrModelsOpenDirectory: + 'settings:document-ocr-models:open-directory', + documentParsingOcrAssets: 'document-parsing:ocr:assets', + documentParsingOcrRequest: 'document-parsing:ocr:request', + documentParsingOcrRespond: 'document-parsing:ocr:respond', + documentParsingOcrCancel: 'document-parsing:ocr:cancel', projectsList: 'projects:list', projectsCreate: 'projects:create', projectsUpdate: 'projects:update',